THREE ESSAYS ON THE TRANSMISSION...

THREE ESSAYS ON THE TRANSMISSION OF

MONETARY POLICY TO NON-BANK

CREDIT ACTIVITY

by

KARL DAVID BOULWARE

ROBERT R. REED, COMMITTEE CHAIRWALTER ENDERS

JUN MASHANE UNDERWOODWILLIAM JACKSON

A DISSERTATION

Submitted in partial fulfillment of the requirementsfor the degree of Doctor of Philosophy in the

Department of Economics, Finance, and Legal Studiesin the Graduate School ofthe University of Alabama

TUSCALOOSA, ALABAMA

2014

1

Copyright Karl David Boulware 2014ALL RIGHTS RESERVED

i

Abstract

This dissertation is composed of three essays that measure the impact of monetary policy

on non-bank credit activity by issuer, composition, and duration. The first essay measures

the dynamic impact of monetary policy on gross repurchase agreement activity of primary

government dealers of the Federal Reserve System. The second essay measures the dynamic

impact of monetary policy on commercial paper activity. The third essay measures the

impact of monetary policy on issuers of asset-backed securities.

In the first essay, we find a positive shock to the federal funds rate significantly affects the

level of credit activity. In particular, repo arrangements longer than a day display persistent

declines. By comparison, overnight financing increases after a delay. This implies that

contractionary monetary policy shocks lead to maturity substitution in the repo market.

Our findings show that credit activity in the repo market is more sensitive to monetary

policy than previously reported in the literature.

In the second essay, our measure of contractionary monetary policy shocks corresponds

to a sharp decline in money market mutual fund assets. Though there is an increase in

aggregate commercial paper volumes, the impact of monetary policy is stronger for issuers

with less liquid balance sheets. Specifically, issuers of asset-backed paper and issuers with

second tier credit ratings. Furthermore, there is evidence of a broad substitution towards

shorter maturities, in particular for asset backed and nonfinancial paper.

ii

In the final essay, we find that an anticipated increase in the target for the federal funds

rate impacts the behavior of ABS issuers. In particular, we find commercial paper issuance

rises while bond issuance falls.

Consequently, our results support the existence of a liquidity risk channel for monetary

policy operating through the total supply of non-bank credit activity. In this manner, our

findings indicate the monetary transmission mechanism contributes to systemic risk in the

shadow banking system through rollover risk. As a result, non-bank credit activity is an

important component of the relationship between monetary policy and financial stability.

iii

Dedication

This dissertation is dedicated to my mother, Dr. Leotta Boulware, for kicking me out of her

house ten years ago, the village that raised me from a child, and to Pesey Kong for preparing

me to be me.

iv

List of Abbreviations, Acronyms, andSymbols

$ American Dollar

= Equal to

∆ First Difference

% Percent

Ω Information Set∑Summation

µ Mean

ρ First-Order Autoregressive Coefficient

σ Standard Deviation

f 0 The Spot-Month Federal Funds Futures Rate

∆r Raw Federal Funds Target Rate Change

∆re Expected Federal Funds Target Rate Change

∆ru Unexpected Federal Funds Target Rate Change

v

ABCPSA Outstanding Asset-Backed Commercial Paper

ABCP Asset-Backed Commercial Paper

ABS Asset-Backed Securities

ADL Augmented Distributed Lag

ALLCPSA Outstanding Commercial Paper

ALLMMMFSA Outstanding Money Market Mutual Fund Assets

DTCC Depository Trust & Clearing Corporation

FINCPSA Outstanding Financial Commercial Paper

FOMC Federal Open Market Committee

FRBNY Federal Reserve Bank of New York

GC General Collateral

GDP Gross Domestic Product

GMAC General Motors Acceptance Corporation

GSE Government-Sponsored Enterprises

IMMMFSA Outstanding Institutional Money Market Mutual Fund Assets

LSAP Large Scale Asset Program

MBS Mortgage Backed Securities

MEP Maturity Extension Program

MF Mutual Fund

Max Maximum Value

Min Minimum Value

MMF Money Mutual Fund

vi

MMMF Money Market Mutual Fund

NBER The National Bureau of Economic Research

NONFINCPSA Outstanding Nonfinancial Commercial Paper

NYMEX New York Mercantile Exchange

OC Overnight and Continuing Agreement

Obs. Number of Observations

p Monetary Policy Instrument

RMMMFSA Outstanding Retail Money Market Mutual Fund Assets

R2 R-squared

SIFMA Securities Industry and Financial Markets Association

SOMA System Open Market Account

SPV Special Purpose Vehicle

TERM Term Agreement

TMPG Treasury Market Practices Group

U.S. United States of America

VAR Vector Autoregression

WTI West Texas Intermediate

vii

Acknowledgments

This dissertation represents all that I know and therefore, a great many people contributed

to its completion.

First, I would like to thank Dr. Robert R. Reed, my dissertation chair, for his visionary

guidance. I would also like to thank Dr. Walter Enders, Dr. Jun Ma, Dr. Shane Underwood,

and Dr. William Jackson for not only agreeing to serve on my dissertation committee but

also for helpful discussions, comments, and suggestions.

Appreciation is expressed to my family, friends, mentors, colleagues, and classmates for

limitless encouragement. I also wish to thank all personnel of the Culverhouse College of

Commerce and Business Administration, the Department of Economics, Finance, and Legal

Studies, the University of Alabama Graduate School, the AEA Summer Program, the Federal

Reserve Bank of Atlanta, the Federal Reserve Bank of San Francisco, the Federal Reserve

Bank of New York, the Federal Reserve Bank of Chicago, the Federal Reserve Bank of

Boston, the Federal Reserve Board of Governors, the NBER Summer Institute, the CSWEP

Summer Fellows Program, the Pipeline Mentoring Program, and the SREB-State Doctoral

Scholars Program.

Lastly, I would like to acknowledge the memories of David Anderson, Leroy Brathwaite,

Dr. Winthrop Jones Boulware, Winthrop Jones Boulware Jr., Lillian M. Collier, Gordon H.

Clem, Phillip Farley, Oswaldo Hales, Billy P. Helms, David Jenkins, Diane Johnson, George

and Lenore Payne, Linwood Spruill, Kenvit Marcus Wilkenson, Kenvit Marcus Wilkenson

Jr., Thomas and Sally Wilson, and Ben and Marge Whitehorn.

viii

Table of Contents

Abstract ii

Dedication iv

List of Abbreviations, Acronyms, and Symbols v

Acknowledgements viii

List of Tables xiii

List of Figures xv

Introduction 1

How Do Money Market Conditions Affect Shadow Banking Activity? Ev-idence From Security Repurchase Agreements 3

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Institutional Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Primary Dealers and the Transmission of Monetary Policy to MoneyMarket Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Net Repo Activity by the Primary Dealers . . . . . . . . . . . . . . . 16

The Composition of Net Financing Activity . . . . . . . . . . . . . . 18

Adding the Monetary Authority to the Securitized Banking Model . . 20

Empirical Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

ix

Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Preliminary Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Modeling the Open Market Desk . . . . . . . . . . . . . . . . . . . . 26

Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Benchmark Response – An Increase in the Cost of Credit . . . . . . . 32

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Alternate Interest Rate Measures of Monetary Policy . . . . . . . . . 34

The Response of Aggregate Financing . . . . . . . . . . . . . . . . . . 35

The Response of Financing Fails . . . . . . . . . . . . . . . . . . . . . 36

The Response of Security Asset Classes . . . . . . . . . . . . . . . . . 37

Repo Activity and the System Open Market Account . . . . . . . . . 38

Choosing the Right Policy Instrument: Evidence from Forecast ErrorVariance Decompositions . . . . . . . . . . . . . . . . . . . . . . . . . 42

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Data Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Monetary Policy and the Non-Bank Financial Sector: A Look at Commer-cial Paper 90

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

Institutional Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

The Market for Commercial Paper . . . . . . . . . . . . . . . . . . . 97

Shadow Banking and the Supply of Credit . . . . . . . . . . . . . . . 98


Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Modeling the Federal Open Market Committee . . . . . . . . . . . . 101

Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

x

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Jobless Claims, Oil, and the Interest Rate Response . . . . . . . . . . 106

Money Market Mutual Fund Responses . . . . . . . . . . . . . . . . . 107

Commercial Paper Responses . . . . . . . . . . . . . . . . . . . . . . 107

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

Responses by Collatarelization . . . . . . . . . . . . . . . . . . . . . . 108

Responses by Credit Rating . . . . . . . . . . . . . . . . . . . . . . . 109

Responses by Intermediation . . . . . . . . . . . . . . . . . . . . . . . 109

Responses by Paper Maturity . . . . . . . . . . . . . . . . . . . . . . 110

Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

Data Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Monetary Policy and the Non-Bank Financial Sector: A Look at Issuersof Asset-Backed Securities 135

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

Institutional Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

Shadow Banking and the Supply of Credit . . . . . . . . . . . . . . . 138


Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

Baseline Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

Decomposing Monetary Policy Actions . . . . . . . . . . . . . . . . . 141

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

The Reaction of Conduit Liabilities to Changes in the Target FederalFunds Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

The Reaction of Conduit Liabilities to Federal Funds Rate Surprises . 143

Robustness Check: The Reaction of Conduit Assets . . . . . . . . . . 143

xi

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

Data Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

Conclusion 159

xii

List of Tables

1.1 The Primary Government Securities Dealers . . . . . . . . . . . . . . . . . . 51

1.2 Descriptive Statistics: July 4, 2001 to January 31, 2007 . . . . . . . . . . . . 52

1.3 Net Repo Borrowing and the Cost of Credit: July 4, 2001 to January 31, 2007 53

1.4 Net Financing and the Cost of Credit: July 4, 2001 to January 31, 2007 . . . 54

1.5 Gross Financing and Changes in the Cost of Credit: July 4, 2001 to January31, 2007 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

1.6 Daily Forecast Regressions of Federal Funds Target Rate Changes: July 4,2001-January 31, 2007 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

1.7 Innovations in The Reserve Market: July 4, 2001-January 31, 2007 . . . . . . 56

1.8 First-Stage Results: εNBR Regressed on Instruments: July 4, 2001-January31, 2007 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

1.9 Estimated Slope of the Supply Function for Nonborrowed Reserves: July 4,2001-January 31, 2007 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

1.10 Variance Decompositions of Gross Repo: July 4, 2001-January 31, 2007 . . . 58

1.11 Variance Decompositions of Securities Out and Settlement Fails: July 4, 2001-January 31, 2007 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

2.1 Descriptive Statistics: January 5, 2001 to February 2, 2007 . . . . . . . . . . 119

3.1 The ABS Issuers Balance Sheet - A Snapshot . . . . . . . . . . . . . . . . . 147

3.2 Descriptive Statistics: 1989 Q2 to 2007 Q3 . . . . . . . . . . . . . . . . . . . 148

3.3 The Impact of Monetary Policy on ABB . . . . . . . . . . . . . . . . . . . . 149

3.4 The Impact of Monetary Policy on ABCP . . . . . . . . . . . . . . . . . . . 150

xiii

3.5 The Impact of Monetary Policy on MIX . . . . . . . . . . . . . . . . . . . . 151

3.6 The Impact of Monetary Policy on ABB . . . . . . . . . . . . . . . . . . . . 152

3.7 The Impact of Monetary Policy on ABCP . . . . . . . . . . . . . . . . . . . 153

3.8 The Impact of Monetary Policy on MIX . . . . . . . . . . . . . . . . . . . . 154

3.9 The Impact of Monetary Policy on ASSETS . . . . . . . . . . . . . . . . . . 155

3.10 The Impact of Monetary Policy on ASSETS . . . . . . . . . . . . . . . . . . 156

xiv

List of Figures

1.1 System Open Market Account Holdings: December 18, 2002-January 31, 2007 60

1.2 Temporary Open Market Operation - Tri-party Repo . . . . . . . . . . . . . 61

1.3 Temporary Open Market Operation - Bilateral Reverse Repo . . . . . . . . . 62

1.4 Gross Repo Activity: July 4, 2001-January 31, 2007 . . . . . . . . . . . . . . 63

1.5 Primary Dealers Net Financing By Maturity: July 4, 2001-January 31, 2007 64

1.6 Primary Dealers Net Financing and Fails By Security Class: July 4, 2001-January 31, 2007 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

1.7 Monetary Policy and Securitized Banking . . . . . . . . . . . . . . . . . . . . 66

1.8 Measures of Real Activity: July 4, 2001-January 31, 2007 . . . . . . . . . . . 67

1.9 The Federal Funds Rate and Federal Funds Target in 2001 and 2004 . . . . . 68

1.10 Response of The Federal Funds Rate to Energy and Employment Shocks . . 69

1.11 The Response of Open Market Operations to Reserve Imbalance Shocks . . . 70

1.12 Responses to a Monetary Policy Shock . . . . . . . . . . . . . . . . . . . . . 71

1.13 Response of Real Activity to a Shock to the Federal Funds Rate . . . . . . . 72

1.14 Response of Repo to Cost of Credit Shocks . . . . . . . . . . . . . . . . . . . 73

1.15 Response of Securities Out to Cost of Credit Shocks . . . . . . . . . . . . . . 74

1.16 Response of Financing Fails to Cost of Credit Shocks . . . . . . . . . . . . . 75

1.17 Response of Securities Out By Collateral Class to Cost of Credit Shocks . . . 76

1.18 Response of Overnight and Continuing Securities Out By Collateral Class toCost of Credit Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

1.19 Response of Term Securities Out By Collateral Class to Cost of Credit Shocks 78

xv

1.20 Response of Financing Fails By Collateral Class to Cost of Credit Shocks . . 79

1.21 Response of Repo to Temporary Open Market Operation Shocks . . . . . . . 80

1.22 Response of Securities Out By Collateral Class to Temporary Open MarketOperation Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

1.23 Response of Overnight and Continuing Securities Out By Collateral Class toTemporary Open Market Operation Shocks . . . . . . . . . . . . . . . . . . . 82

1.24 Responses of Term Securities Out By Collateral Class to Temporary OpenMarket Operation Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

1.25 Response of Financing Fails By Collateral Class to Temporary Open MarketOperation Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

1.26 Response of Repo to Permanent Open Market Operation Shocks . . . . . . . 85

1.27 Response of Securities Out By Collateral Class to Permanent Open MarketOperation Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

1.28 Response of Overnight and Continuing Securities Out By Collateral Class toPermanent Open Market Operation Shocks . . . . . . . . . . . . . . . . . . . 87

1.29 Response of Term Securities Out By Collateral Class to Permanent OpenMarket Operation Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

1.30 Response of Financing Fails By Collateral Class to Permanent Open MarketOperation Shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

2.1 Commercial Paper Outstanding . . . . . . . . . . . . . . . . . . . . . . . . . 120

2.2 Money Fund Assets Outstanding . . . . . . . . . . . . . . . . . . . . . . . . 121

2.3 Commercial Paper and the Supply of Credit . . . . . . . . . . . . . . . . . . 122

2.4 Information Arrival for week = w . . . . . . . . . . . . . . . . . . . . . . . . 123

2.5 FOMC Reaction Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

2.6 Responses of Jobless Claims, Oil, and the Interest Rate . . . . . . . . . . . . 125

2.7 Responses of Money Market Mutual Fund Assets . . . . . . . . . . . . . . . 126

2.8 Responses of Commercial Paper Outstanding . . . . . . . . . . . . . . . . . . 127

2.9 Responses by Collateralization . . . . . . . . . . . . . . . . . . . . . . . . . . 128

2.10 Responses by Credit Rating . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

xvi

2.11 Responses by Intermediation . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

2.12 Responses of Issuance by Maturity - All . . . . . . . . . . . . . . . . . . . . 131

2.13 Responses of Issuance by Maturity - Asset-backed . . . . . . . . . . . . . . . 132

2.14 Responses of Issuance by Maturity - Financial . . . . . . . . . . . . . . . . . 133

2.15 Responses of Issuance by Maturity - Nonfinancial . . . . . . . . . . . . . . . 134

3.1 Assets and Liabilities of Asset-Backed Issuers Over Time . . . . . . . . . . . 157

3.2 Securitization and the Supply of Credit . . . . . . . . . . . . . . . . . . . . . 158

xvii

Introduction

This dissertation is composed of three essays that study the transmission of monetary policy

to non-bank credit activity. From 1945 to 2008, term debt markets played an increased

role in the supply of credit and as a result, non-bank financial intermediaries have become

important sources of credit. Furthermore, money market instruments and not deposits are

the main source of funds for the financial sector at the margin. As the initial stages of the

monetary transmission mechanism generally take place through money markets, the actions

of the central bank can impact activity in these credit markets in systematically important

ways.

The first essay, “How Do Money Market Conditions Affect Shadow Banking Activity?

Evidence From Security Repurchase Agreements,” focuses on monetary policy’s impact on

gross repo activity by the primary government securities dealers. The repo market was

one of two money markets severely disrupted during the financial crisis and is the main

funding market of broker dealer investment banks. This work fits into an emerging body of

literature on how monetary policy affects risky behavior in financial markets. While Bech

et al. (2012) and D’Amico et al. (2013) have studied monetary policy’s impact on repo

prices, we find supporting evidence for a monetary transmission channel, operating through

the total volume of repurchase agreements.

The second essay, “Monetary Policy and the Non-Bank Financial Sector: A Look at

Commercial Paper,” studies monetary policy’s impact on the market for commercial paper.

At the end of 2006, outstanding commercial paper was the largest of all outstanding money

1

market instruments. During the financial crisis the market for commercial paper was severely

disrupted thereby impacting the allocation of real economic investment. Some authors have

studied commercial paper activity and monetary policy, notably Kashyap, Stein, and Wilcox

(1993), but none have done so in the risk channel framework. In particular, our analysis

incorporates money market mutual funds, the largest investors in commercial paper. We find

our results support the existence of a liquidity risk channel for monetary policy operating

through the total supply of commercial paper.

The last essay, “Monetary Policy and the Non-Bank Financial Sector: A Look at Issuers

of Asset-Backed Securities,” looks at a subset of the non-bank financial system, or issuers of

asset-backed securities. Many financial institutions sponsor pass-through vehicles, or special

purpose vehicles (SPVs), as a way to reduce financing costs. However, the gains in efficiency

by engaging in off-balance sheet activity also come with the potential loss of stability due

to balance sheet fragility. As a result, the balance sheet composition of these vehicles, while

simple, is a key determinant in the level of systemic risk in the financial system. We use

the asset-backed issuers’ simple liability structure to provide evidence on the existence of a

risk-taking channel of monetary policy. We find that an anticipated increase in the target

for the federal funds rate impacts the behavior of ABS issuers. Notably, by increasing the

balance sheet duration gap between long term assets and short term liabilities.

2

Chapter 1

How Does Monetary Policy AffectShadow Banking Activity? EvidenceFrom Security RepurchaseAgreements

I Introduction

A repurchase agreement (or ‘repo’) is a collateralized money market loan used often in the

process of intermediation in the shadow banking sector.1 (Financial Crisis Inquiry Commis-

sion, 2011) As described by Geithner (2008), overnight tri-party repos funded approximately

$2.5 trillion of assets in early 2007. Prior to the recent financial crisis, along with federal

funds, repurchase agreements represented the largest net increase in liabilities of the U.S.

financial sector. Moreover, during the crisis, they represented the largest net decrease.2

In contrast to standard secured lending arrangements, borrowers in the repo market sell

bonds with the agreement to buy them back at a fixed price at a forward date. The difference

1According to Bernanke (2010), “Shadow banks are financial entities other than regulated depositoryinstitutions (commercial banks, thrifts, and credit unions) that serve as intermediaries to channel savingsand investment...Leading up to the crisis, the shadow banking system, as well as some of the largest globalbanks, had become dependent on various forms of short-term wholesale funding. Over the past 50 years orso, a number of forms of such funding have emerged, including, commercial paper, repurchase agreements(repos), ...., and others. In the years immediately before the crisis, some of these forms of funding grewespecially rapidly; for example, repo liabilities of U.S. broker dealers increased by 2-1/2 times in the fouryears before the crisis.”

2See Woodford (2010) for details.

3

between the sale and repurchase price is called the repo rate and reflects interest on the loan.

The securities sold act as collateral in the cash-loan transaction. During the term of the loan,

lenders utilize the securities as their own.3 In so doing, both parties seek to minimize the

cost of default. (Mills and Reed, 2008)

Naturally, securities dealers rely on the repo market to finance intermediation in securities

markets. However, relying on repos for short-term liquidity remains problematic.4 In fact, in

2009 the Task Force on Tri-Party Repo Infrastructure concluded: “repo arrangements were

at the center of liquidity pressures faced by securities firms at the height of the financial

crisis.”5

Moreover, institutions active in the repo market generally do not have access to the

discount window and resort to “fire sales” in order to raise funds quickly.6 As a result, the

absence of a lender of last resort exacerbates problems in the repo market. (Martin et al.,

2013) As stressed by Gorton and Metrick (2012), instabilities in the repo market can cascade

into the real economy by shutting off debt market intermediation.7 As the initial stages of

the monetary policy transmission mechanism generally take place through money markets,

the actions of the central bank can impact the repo market in systemically important ways.8

Thus, it is important to understand the transmission mechanism of monetary policy through

the repo market. (Kohn, 2008)

Unfortunately, inadequate data complicates any empirical analysis of repo activity.9

3As outlined in the Bankruptcy Amendments and Federal Judgeship Act of 1984.4Notably, Copeland et al. (2012b) contends that major reforms are necessary to improve stability in the

tri-party repo market.5Bernanke (2010) describes how institutions experienced funding disruptions in the repo market during

the crisis. Traders call these temporary events “liquidity holes,” an outcome akin to a bank run. See Taleb(1997, p. 69) and Morris and Shin (2004).

6For further discussion see Begalle et al. (2013).7See also the discussion by the Task Force on Tri-party Repo Infrastructure (2012) about the early

morning unwind of all tri-party repo transactions.8Bernanke (2007) points out monetary policy may contribute to risks in the global financial system by

operating through the money markets and thereby the nonbank financial institutions who rely on them. Forexample, in the four years prior to the summer of 2007, the target rate for federal funds increased by 4.25percentage points (425 basis points). During the same sample, repo liabilities of US broker dealers increasedby 250 percent. (Bernanke 2010)

9See Adrian et al. (2013) for a summary of available data sources.

4

(Bernanke, 2012) However, the primary dealers of the Federal Reserve file the FR 2004

Reports with the Federal Reserve Bank of New York (FRBNY), as required by law, on an

ongoing basis. The FRBNY publishes the FR 2004 Reports weekly and the FR 2004C in-

cludes financing transactions classified as repurchase agreements by maturity. Furthermore,

the report encompasses dealer financing by security type, maturity length, and financing

fails. Therefore, the FR 2004C report is not only a high frequency measure of repo activity,

but it also provides contract details nonexistent in other data sets.10

As we describe in more detail below, there are two important ways the FR 2004C report

measures financing flows in the shadow banking system. First, the data is collected from

the primary dealers, an influential group of intermediaries in the shadow banking system.11

Second, the dealers report financing on a gross basis. As explained by Krishnamurthy et al.

(2013), the volume of gross flows contribute to the probability that defaults feed through

the shadow banking system across dealers.12 In this manner, the FR 2004C reflects the level

of systemic risk in the shadow banking system. As a result, studying the high frequency

response of dealer financing to monetary conditions gives policymakers insight about how

their actions affect conditions in the shadow banking system over time.

In addition to regularly reporting repo transactions, primary dealers serve as on-demand

counterparties for the FRBNY’s trading desk. Consequently, the initial stages in the exe-

cution of monetary policy directly transmit to primary dealers. Yet, their role as monetary

propagation mechanisms continues to be ignored – even though primary dealers occupy the

nexus of money creation and shadow credit facilitation. Hence, the FR 2004 reports provide

a distinct view of the monetary transmission mechanism operating through repo market

activity.

The objective of this paper is to provide empirical evidence on a transmission mechanism

10It also avoids measurement error found in the Flow of Funds Accounts. (Krishnamurthy and Nagel,2013)

11See Pozsar et al. (2010) for an overview of the shadow banking system.12FR 2004 data represents market volumes, not origination’s. Therefore, multiple transactions in the same

security is possible.

5

for monetary policy to shadow banking activity which operates through the total volume

of repurchase agreements. The number of empirical studies measuring the effect of mon-

etary policy on repo activity is limited though Adrian and Shin (2008) find that changes

in the federal funds rate are negatively correlated with the growth of repo financing by

dealer banks.13 In comparison to their single equation analysis, we analyze the response

of dealer repo activity through impulse response functions of an estimated vector autore-

gression (VAR) model. As explained by Stock and Watson (2001), single equation analysis

has a number of limitations relative to VARs. Thus, our VAR analysis produces a number

of insights that are difficult to determine in single equation analysis such as the dynamic

effects of money market conditions which feed through macroeconomic conditions to repo

market activity. Consequently, the dynamic effects of monetary shocks can be more precisely

estimated as the indirect effects through macroeconomic variables are taken into account.

Furthermore, our main identification restriction supports the use of a recursive VAR model

which allows us to characterize the monetary transmission mechanism through the repo mar-

ket.14 In contrast to Adrian and Shin, we find that the coefficient estimates for the effects

of policy are twice as large. Thus, our benchmark results provide further empirical evidence

that policymakers need to anticipate financial conditions in the repo market when evaluating

monetary transmission mechanisms. In this manner, our findings indicate there are strong

connections between the design of monetary policy and macroprudential policy in addition

to the standard concerns about real activity and price stability.

In addition to gross repo activity, we consider whether the maturity structure of repo

financing responds uniformly to shocks to the federal funds rate. A thorough study of the

maturity structure is necessary as many have argued that dealer banks increasingly relied

on overnight repo for funding prior to the 2007 crisis. One hypothesis is that the shift

13Adrian and Shin (2008) also find strong correlations between the growth of repo financing by dealerbanks, dealer bank asset growth, and future real macroeconomic activity. In related work, Adrian and Shin(2010) adopt the same estimation strategy but use changes in the target federal funds rate over a one weekperiod in lieu of the effective rate over a thirteen week period.

14As argued by Bernanke and Blinder (1992) and Bernanke and Mihov (1998), the identification assumptionthat monetary policy affects the economy with a lag is more realistic at the weekly frequency.

6

towards overnight funding was due to increasing demand for money market instruments.15

However, our evidence indicates that the monetary tightening between 2004 and 2006 may

have played a role – while an increase in the federal funds rate is followed by a reduction

in overall repo activity, the decline is concentrated in agreements with maturities longer

than a day. In contrast, we find a delayed positive response in overnight repos backed by

Treasury securities. Therefore, higher levels of the federal funds rate were partially offset

with overnight funding using collateral of the highest quality. As a result, we conclude the

“flight to maturity” in the repo market was also due to higher money market interest rates.

Consequently, monetary policy likely contributed to systemic risk in the shadow banking

system during the run-up to the crisis.

In light of the recent debate on the use of unconventional monetary policy tools, we

also study the impact of various components of the Federal Reserve’s balance sheet. As

communicated by the Federal Reserve, Large Scale Asset Programs (LSAPs) were introduced

to ease financial conditions and promote credit activity. Consequently, we study the impact of

the size of the Federal Reserve’s Treasury portfolio to gain insights into how LSAP programs

can affect activity in the repo market. In comparison to the unprecedented size of purchases

of Treasuries in recent years, standard open market operations are designed to minimally

impact the price of these securities. However, we find that the resulting reduction in the

public’s Treasury stock does affect repo market activity. Consistent with Bartolini et al.

(2011)’s ranking of collateral values by security classes, we find persistent increases in the

use of agency securities for overnight financing. For maturities longer than a day, we find

evidence of a substitution towards corporate securities. While the motivation for LSAPs is

to ease financial conditions, there is some evidence that they may contribute to risk in the

repo market. For example, in response to the withdrawal of Treasury bills in standard open

market operations, there is an increase in fails of mortgage backed securities (MBS). Thus,

one can infer that that the large volumes of purchases of longer-dated securities in a LSAP

15Brunnermier (2009) is perhaps the most well known to make this argument. By comparison, Gorton etal. (2012) present evidence that the “flight to maturity” actually started in July 2007.

7

would have similar but exaggerated effects on repo activity. Furthermore, the intensity of

this transmission channel is related to the fraction of Treasury Bills to Treasury securities

held by the System Open Market Account (SOMA) and as a result, informative about the

effects of the Maturity Extension Program, also known as ‘Operation Twist.’16

In addition to outright open market operations, the Federal Reserve has increasingly

turned to the repo market when accommodating daily variations in the supply of reserves.

Fed repos are one type of temporary open market operation used most often to control the

level of bank reserves. Under a Fed repo, the open market desk temporarily lends funds

to primary dealers and accepts general collateral (GC) in Treasuries, agency, and MBS. By

temporarily raising demand for GC, the open market desk promotes the ‘collateral rights’ of

holders of GC securities. In turn, such transactions have a delayed but persistent positive

impact on repo activity with maturities longer than a day. Moreover, the withdrawal of

Treasury, agency, and MBS leads to a substitution towards overnight arrangements backed

by corporate securities. In contrast to outright open market operations, an increase in Fed

repos corresponds to a persistent decrease in financing fails. Furthermore, the decrease in

financing fails is the largest and most persistent of all fail reactions by policy instrument.17

In comparison to Fed repos, Fed reverse-repos have recently attracted attention as the

Federal Reserve considers expanding their use in order to improve control over money market

rates. Such a tool would seek to “...reassure investors that the Federal Reserve has sufficient

tools to manage monetary policy effectively even with a very large balance sheet.” (Dudley,

2013) While Fed repos remove the supply of assets that serve as collateral, Fed reverse-repos

increase the supply of securities used as collateral. We find that this component of the Fed-

eral Reserve’s balance sheet also affects activity in the repo market. While Fed repos appear

16On September 21, 2011 the Federal Reserve announced the Maturity Extension Program and Reinvest-ment Policy for the purpose of extending the average maturity of Treasury securities in the SOMA portfolio.This program was subsequently continued on June 20, 2012. (Board of Governors of the Federal ReserveSystem, 2011)

17Garbade et al. (2010) discuss the evolution of the recently introduced Treasury Market Practices Group(TMPG) fails charge which was implemented in May 2009 and designed to to encourage timely settlementfor Treasury securities.

8

to promote long-term financing, reverse-repos lead to maturity substitution by persistently

decreasing term agreements and increasing overnight arrangements. By increasing the sup-

ply of assets that are readily accepted as collateral in the repo market, there is a temporary

decrease in financing fails for Treasury securities. However, as the relative supply of agency

securities is lower after a reverse-repo, there is a persistent increase in financing fails for

agency securities. Thus, our results suggest that the Federal Reserve faces a trade-off be-

tween temporarily promoting stability in the repo market versus increasing fails via agency

securities. We conjecture one reason this trade-off exists is due to the manner each open

market operation is conducted. For example, the transmission mechanism for Fed repos

operates through a wide array of security classes. In contrast, reverse-repos are regularly

backed by Treasury bills.

Our results make important contributions to an emerging literature on shadow banking

activity. The closest paper to our work is Bech, Klee, and Stebunovs (2011) who primarily

study the instantaneous impact of monetary policy on repo rates. Notably, they identify

periods in which the repo rate does not adjust in line with changes to the federal funds

rate. This indicates that repo markets may be prone to inefficiency. In contrast to our work,

Bech et al. study the overnight Treasury general collateral repo rate at the daily frequency.

However, the focus of our research is on the impact of monetary shocks on the level of repo

activity rather than rates alone. In this manner, we show that policy shocks affect the

amount of credit activity among key participants in the shadow banking system. Moreover,

we observe that monetary policy not only affects the maturity structure in repo markets,

but also the collateral used.

In comparison to the effects of outside money on repo market activity, Sunderam (2012)

argues that short-term liabilities of shadow banking institutions (including repurchase agree-

ments) respond to money demand. That is, Sunderam’s work emphasizes that the supply

of inside money to the financial system was a response to changes taking place internally

within the financial system for high-powered (‘inside’) money. However, our focus is on the

9

endogenous response of shadow banking institutions to changes in the availability of outside

money through the central bank.

Other relevant work addresses the stability of repo markets and its systemic importance.

Notably, Gorton and Metrick (2012) study price data on bilateral repos and show evidence

supporting the hypothesis that increases in haircuts on private label asset backed securities

(ABS) started runs in the system. By comparison, Krishnamurthy et al. (2013) measures

the size and composition of loans by the twenty largest lenders in the repo market just

prior to, during, and after the crisis. Their data shows that repo financing to the shadow

banking system was backed mostly by Treasury, agency, and corporate securities rather than

private label ABS. In addition, Krishnamurthy et al. (2013) finds evidence that dealers’

balance sheets are a systemically important channel for shadow banking activity operating

through the perceived riskiness or illiquidity of the collateral pledged. Tri-party lenders such

as money market mutual funds and securities lenders abruptly stop lending after a perceived

increase whereas bilateral lenders such as other dealers and hedgefunds demand higher levels

of collateralization. Copeland et al. (2012a) measures daily activity in the tri-party repo

market during the summer of 2008 through early 2010 and come to a similar conclusion.

They find that in contrast to the bilateral repo market, funding and haircuts in the tri-party

market were stable with the exception of Lehman Brothers’ bankruptcy. Moreover, this

change in the willingness to lend to dealers after Lehman’s failure was not gradual.

The remainder of the paper is organized as follows. Section 2 proceeds by outlining insti-

tutional details linking the transmission of monetary policy, the role of primary dealers, and

the repurchase agreement market. Section 3 describes the data and empirical methodology.

Section 4 reports the benchmark results. Section 5 discusses the robustness of the benchmark

results and extends our analysis to include policy instruments in the System Open Market

Account. Section 6 concludes.

10

II Institutional Details

Primary dealers inhabit a special place in the U.S. financial system. If monetary policy

influences dealer financing activity in the repo market unfavorably, then monetary policy

directly contributes to financial instability in the shadow banking system. In the following

section we discuss the main institutional features that link the transmission of monetary

policy to shadow banking activity.

II.1 Primary Dealers and the Transmission of Monetary Policy toMoney Market Activity

Primary dealers denote a key subgroup of securities dealers.18 Table 1 lists the all of the

primary dealers during our sample which includes bank subsidiaries and stand alone broker-

dealers. The main role of primary dealers in the Federal Reserve System is to act as the

on-demand trading counterparty of the FRBNY in its implementation of monetary policy.

There are two ways that this occurs. The first mechanism is through permanent open market

operations while the second is “temporary,” occurring on a daily basis.

The traditional interpretation of open market operations – the method of implementing

changes in monetary policy – occurs through permanent open market operations. If the

Fed wants to lower the target federal funds rate, it purchases Treasury securities from the

primary dealers which increases the supply of reserves to the financial system. Purchases are

generally held to maturity. By comparison, it sells securities to increase the target. These

transactions affect the amount of assets held outright in the SOMA domestic portfolio.

Such transactions take place via the open market desk at the FRBNY following instructions

by the Federal Open Market Committee (FOMC). The top left panel of Figure 1 plots

the total amount of SOMA securities bought outright from December 18, 2002 to January

31, 2007. On average, 100% of the securities purchased are Treasuries with close to 60%

18In the third quarter of 2012, primary dealers underwrote 79, 70, and 49 percent of all US, US agency,and corporate bonds. (Bloomberg Global Fixed Income League Tables, 2012)

11

of the portfolio invested in notes and bonds. Over time, the portfolio holdings steadily

grow, consistent with permanent operations being used predominantly to offset increases in

circulating currency. The maturity distribution of the portfolio is plotted in the top right

panel. Securities with maturities ranging from 16 days to 5 years represent, on average,

about 60% of investments. Interestingly, outright holdings seem to be balanced between

short and medium term maturities. While the largest average holdings mature between one

and five years, Treasury bills average about 40% of the portfolio.

In addition to permanent changes in reserve balances, the supply and demand for re-

serves in the federal funds market fluctuates daily.19 Changes in demand take place for a

number of reasons. Notably, Hamilton (1997) describes how the amount of Treasury deposits

at Federal Reserve banks varies on a daily basis depending on fiscal receipts and expendi-

tures.20 To offset these fluctuations, the Federal Reserve also conducts “temporary” open

market operations through repurchase agreements and reverse repurchase agreements.21 In

a temporary open market operation that adds liquidity to the banking system, the desk at

the FRBNY lends funds to a primary dealer in exchange for collateral. As illustrated in

Figure 2, delivery of collateral to the lender is settled as “tri-party” through the custodial

account of the FRBNY’s clearing bank. By comparison, Figure 3 depicts how reverse repos

take place as bilateral transactions, settle delivery versus payment, and may also take place

through institutions other than primary dealers.22

19This is by design. The Federal Reserve System operates at a ‘structural deficiency’ meaning reservesadded permanently are less than the amount needed. This enhances the Desk’s ability to control liquidityslack in the system.

20Reserves can also fluctuate because of float and currency held by the public.21The Master Repurchase Agreement published by the Securities Industry and Financial Market Associa-

tion is the base form documenting the legal terms and conditions under which the FRBNY and counterpartiesmay undertake repo market transactions. Federal Reserve Banks have engaged in repurchase agreementssince 1916 and switched from matched sale purchase to reverse repurchase agreements on December 13, 2002.(Simmons, 1954)

22Prior to the recent extension of counterparties for reverse-repos, such transactions took place as bilateralarrangements rather than tri-party as the Federal Reserve was not considered to be a dealer in clearing banks’systems. Consequently, reverse repos with the primary dealers were bilateral transactions. We thank AntoineMartin for clarifying these details with us. In addition to the primary dealers, there were also reverse-reposconducted through foreign official and international accounts. As of August 18, 2010 the Fed has expandedits list of eligible participants to include Money Market Mutual Funds (MMF), Government-SponsoredEnterprises (GSE), and banks.

12

The bottom panel of Figure 1 plots SOMA holdings of repo and reverse repo. Temporary

open market operations represent on average less than 7% of the SOMA portfolio and are

dominated by transactions with maturities of 13 days or less. Consistent with their daily

policy function, Fed repos show more weekly variation than reverse repos. Finally, excluding

the negligible fraction of repos maturing in over 13 days, temporary open market operations

generally occur no later than 9:30 am Eastern time each morning. Permanent open market

operations, on the other hand, occur anytime after the morning auction. (Federal Reserve

Bank of New York, 2002)

In addition to trading with the open market desk, primary dealers are expected to reg-

ularly participate in all U.S. government debt auctions. Bids submitted by the primary

dealers are competitive and as such determine the yield and price paid by filled competitive

and noncompetitive bids for each Treasury issue.23 Dealers then hold or trade any securities

awarded. Thus, “primary dealer” is short for primary dealer in U.S. government securities.

The Weekly Report of Dealer Financing and Fails

As primary dealers are significant intermediaries in money markets, their activity provides

important information to the Federal Reserve for federal funds targeting. Consequently,

primary dealers are required to file form FR 2004 on an ongoing basis. It is collected,

consolidated, and released publicly every week by the FRBNY. The FOMC uses the report

to monitor the condition of the U.S. Treasury securities market which allows it to carry out

more informed open market operations and actions as fiscal agent of the U. S. Treasury.

The forms include the Weekly Report of Dealer Positions, the Weekly Report of Cumulative

Dealer Transactions, the Weekly Report of Dealer Financing and Fails, the Weekly Report of

Specific Issues, the Daily Report of Specific Issues, and the Daily Report of Dealer Activity

in Treasury Financing.24

23Except for the 10 year note, auctions for specific issue Treasury bills and bonds occur on weekly andmonthly schedules respectively.

24Reporting guidelines for preparing the FR 2004 primary government securities dealers reports can befound at: http://www.federalreserve.gov/reportforms.

13

http://www.federalreserve.gov/reportforms.

The weekly releases show market data reported by the legal entity that functions as the

primary dealer on outright positions, cumulative transactions, gross financing, and cumula-

tive fails in U.S. Treasury, government agency, agency MBS, and corporate debt securities.

These four security classes represent the four largest classes of collateral in the tri-party repo

market.25 (Copeland et al. 2012a) Reporting is as of the close of business each Wednesday

and the FRBNY releases summary data each Thursday after market hours. The FRBNY

staff reviews data submitted on the FR 2004 reports and, as needed, may ask for explana-

tions or revisions. However, other than monitoring their participation in these requirements,

the FRBNY has no regulatory power over the primary dealers. Thus, the FRBNY expects

the dealers to submit accurate data but does not audit it.

The Weekly Report of Dealer Financing and Fails, or FR 2004C, collects outstanding

financing arrangements and fails for the calendar week. Fails are reported on a cumulative

basis for the reporting period for both lending and borrowing arrangements. The amount

reported is the transaction’s principal value on the day the failed trade was to be settled. Fi-

nancing data is reported on a gross basis of actual funds paid or received. It is disaggregated

by security class and maturity length. Repo activity is identified as a subset of aggregate

financing activity and disaggregated only by maturity. Overnight and continuing contracts

(OC) mature after one business day and can be renewed daily unless terminated by either

party. Term agreements (TERM) have a specified length of more than one day.

Previous work has emphasized the importance of repo market activity by primary dealers

in the extension of credit to the real economy. For example, Gorton and Metrick (2012) stress

that repo financing plays a significant role in “securitized banking” in which dealers use repo

to purchase large volumes of loan obligations and make markets in the repackaged tranches

of asset-backed securities. In particular, King (2008) points out that in the financial quarter

prior to Bear Stearns’ failure, five ‘pure’ investment banks funded between 28 and 55 percent

25The data collected by Krishnamurthy et al. (2013) does not distinguish between agency and agencyMBS.

14

of the financial instruments on their balance sheets through repurchase agreements.26 In

addition, at its peak, the largest dealer position in the tri-party repo market totaled over

$400 billion dollars. (Federal Reserve Bank of New York, 2010) Furthermore, repos can

be collateralized with assets dealers do not hold outright. For example, collateral held by

the dealer when executing the first leg of a reverse repo is often reused for additional repo

funding, a process also known as “re-hypothecation.”27

Figure 4 plots our three main repo variables and their financing analogs: gross repo,

gross overnight and continuing repo, and gross term repo. As Krishnamurthy et al. (2013)

explains, the inclusion of inter-dealer repo in the FR 2004C eliminates gross repo as a useful

measure of unique repo flows into the shadow banking system. However, for our purposes,

gross repo is an advantageous quantity to use for a few key reasons. First, because one

security is often needed by dealers to satisfy more than one short position, gross repo mea-

sures the velocity of repo collateral which is analogous to a money multiplier. (Gorton and

Metrick, 2012) Second, gross repo is overwhelmingly composed of inter-dealer repo. Al-

though inter-dealer repo is not an original funding source for the shadow banking system,

it does measure the reallocation of original funding into illiquid and risky securities. (Kr-

ishnamurthy et al., 2013) Third, if the federal funds rate is cointegrated with the GC repo

rate and this spread is a strong determinant in the quality of collateral composition then

it is important to understand how monetary policy transmits not only to original funding

but also its reallocation. (Bech et al., 2011; Bartolini et al., 2010) Lastly, during a run,

repo lenders are unwilling to lend against illiquid and or risky collateral. (Copeland et al.,

2012a; Krishnamurthy et al., 2013) Furthermore, lenders have no incentive to return high

quality collateral to institutions experiencing runs. Therefore, if the amount of short interest

exceeds the total quantity of the security issued during a run, it can be too burdensome for

26At the time, the five investment banks of Bear Stearns, Goldman Sachs, Lehman Brothers, Merrill Lynch,and Morgan Stanley all had primary dealer status.

27Re-hypothecation occurs when a dealer secures financing using collateral posted by a client. (Duffy,2010) For example, Krishnamurthy et al. (2013) finds that gross repo activity from the fourth quarter of2006 to the first quarter of 2010 was on average, 4 times as large as the amount lent by MMF’s and SL’s.

15

borrowers to acquire the specific collateral needed to avoid distress. Combined with evidence

by Krishnamurthy et al. (2013) that dealers continue lending against illiquid and risky col-

lateral even during tumultuous money market conditions, gross repo can proxy for the risk

of systemic contagion in the shadow banking system. As a result, Adrian and Shin (2009)

among others, have argued gross repo may be a better measure of the financial system’s

health than traditional monetary aggregates.

II.2 Net Repo Activity by the Primary Dealers

In order to better understand the primary dealers’ economic use of repo over time, we graph

two simple measures of dealer borrowing: net financing and net repo financing. Net repo

financing is a measure of the net amount of borrowing by primary dealers using repurchase

agreements. It is calculated by subtracting the difference between gross repo (out) and

reverse repo (in) thereby netting out the effects of re-hypothecation. However, net repo

financing does not account for collateralized securities lending which is economically identical

to repos but under generally accepted accounting principles not always treated as such.28

We therefore include net financing, the net amount of funds borrowed by primary dealers

in all reported financing transactions. Net financing is calculated as the difference between

securities delivered (out) minus securities received (in). The bottom left panel of Figure 5

plots the level of net repo and net repo financing by maturity during our sample period.

The top and bottom lines in the figure represent net overnight and term repo borrowing

respectively. Moving clockwise by panel shows time-series plots for each of our three measures

of net repo financing with its net financing analog.

There are a number of observations that one can make from looking at Figure 5. First, and

consistent with previous findings, throughout our sample, primary dealers are net borrowers

in repo markets. Furthermore, borrowing occurs through overnight arrangements. However,

the dealers are net lenders in term activity. Consequently, like traditional intermediaries,

28This is one source of the measurement error found by Krishnamurthy and Nagel (2013) in the Flow ofFunds Accounts.

16

primary dealers engage in maturity mismatch in that most of the borrowing takes place

overnight but lending is for longer periods.

Over time, the level of net borrowing activity varies. During the second half of the

sample, net repo financing grew at a faster pace than in the first half. Gorton and Metrick

(2012) observe that dealers’ reliance on the repo market is positively related to growth in

the securitized banking system. However, the acceleration in repo borrowing by primary

dealers appears to also coincide with a monetary tightening cycle beginning June 30, 2004.

As stressed by Woodford (2010), the increase in net repo could be the by-product of the

compression in term spreads that began prior to but continued during the FOMC’s 2004 to

2006 tightening cycle. Further evidence from bank loan performance and mortgage growth

to subprime ZIP codes is strongly supportive of this hypothesis. (see Federal Reserve Bank

of New York (2013) and Mian and Sufi (2009).) On the other hand, term lending shows no

such evidence of demand effects. Term lending is stable prior to the summer of 2004, but

increases abruptly in November of that year, and eventually bottoms out around one year

later.

When we compare primary dealers’ net repo borrowing with net financing, a slightly

different picture emerges. Dealers are still net borrowers engaged in maturity mismatch.

However, the top left panel of Figure 5 shows that dealers’ wholesale funding needs did not

change over the sample. Furthermore, as plotted in the two right panels, the intensity of

overnight borrowing and term lending is different for the two measures. Net repo dominates

overnight agreements and net financing dominates term.

One possible explanation for the discrepancy is measurement error. As shown by Krish-

namurthy and Nagel (2013), the Flow of Funds overstates the amount of net repo borrowing

of commercial banks because it does not include repo positions of securities lenders. A com-

parison of the time series properties of securities out, securities in, and gross reverse repos

reveals what could be a similar overstatement in the FR 2004C. The bottom left panel of

Figure 4 shows securities out and securities in are highly correlated but securities in and

17

gross reverse repos are not. Consequently, it seems that net repo borrowing overstates net

financing by the primary dealers. However, another possible explanation is regulatory ar-

bitrage. Varying degrees of balance sheet ‘window dressing’ can be beneficial for non-bank

intermediaries and their hedge fund clients. Upon inspection of the weeks surrounding the

end of financial quarters, Figure 4 is suggestive of dealers engaging in this behavior. In

particular, with term agreements more than overnight. One well known example of dealer

window dressing is Lehman Brothers’ use of repo 105. Once recognized as a true sale under

GAAP, repo 105 removed assets from Lehman’s balance sheet thereby lowering the amount

of leverage disclosed to the public. (Valukas, 2010) On the other hand, if a client wishes to

avoid recording assets on their balance sheet, a dealer will instead engage in a collateralized

loan.

II.3 The Composition of Net Financing Activity

As argued by Gorton and Metrick (2012) and documented extensively by Krishnamurthy et

al. (2013) and Copeland et al. (2012a), repo markets appeared to breakdown during the

crisis because money market funds and securities lenders became unwilling to lend against

private label MBS. In more general market conditions, one can extrapolate that institutions

would be less willing to hold more credit-sensitive and illiquid assets as money markets

tighten, particularly agency MBS and non-fed eligible collateral such as corporate bonds.

For example Bartolini et al. (2010) find that classes of securities can be ranked by their

collateral values in the general collateral market. As a way of generating insights into these

issues, we look at primary dealer net financing by security class.

The top left panel of Figure 6 plots net borrowing among dealers by the type of collateral

used. Consistent with Copeland et al. (2012a)’s tri-party data, the dominant form of

collateral is agency MBS. Notably, its importance slowly grows over the sample period. In

this manner, our data are consistent with Krishnamurthy et al. (2013) who modify the

arguments of Gorton and Metrick (2012) to recognize that private label MBS was not the

18

dominant form of repo collateral in the securitized banking model. Over most of our sample,

both agency and corporate securities oscillate between second place. However, in 2005 net

borrowing in corporate and agency securities diverge and corporates take over. On the other

hand, net lending by dealers is dominated by Treasuries.

The aggregate data provides important insights, lending further support to Woodford

(2010)’s arguments about demand effects. However, the aggregate time-series properties of

collateral used may vary by maturity. The top right panel of Figure 6 looks at net collateral

used in overnight agreements. By comparison, the bottom right panel of Figure 6 plots

the securities used as collateral in term arrangements. When looking at net financing by

collateral and maturity the only evidence for an increase in demand is the steady increase in

net overnight borrowing in MBS over our sample. Net overnight borrowing using corporates

increases steadily beginning in 2004 whereas overnight borrowing in the other two security

classes do not show much variation. Furthermore, we do not find evidence that lenders

distinguish collateral in a predictable manner by collateral type when the FOMC tightens.

In contrast, we do find evidence of collateral rankings in term financing. Net term financing

shows strong variation in lending and strong declines after the 2004 policy tightening. Only

Treasuries show a consistent increase in net term lending over our whole sample. The other

three classes show gradual increases starting with corporates in 2003, MBS in 2004, and

agency in 2006.

Liquidity Disruptions

While repurchase agreements are specified to be settled at a particular time and date, occa-

sionally, liquidity problems emerge. For example, sometimes a participating party may fail

to deliver or receive a security on time. The bottom left panel of Figure 6 is a time-series

graph of cumulative fails by security class. There are some notable observations to take

from the data. First, there are several episodes of major dislocations that occur. In fact,

there are fails in every week. The biggest failures occur in Treasuries, followed by MBS,

19

and agency securities. Moreover, failures appear to be persistent events. Consequently, fail

persistence is likely to increase counterparty risk and reduce market liquidity if a number

of market participants are relying on receipt of the same class of securities throughout the

same trading period.

Fleming and Garbade (2005) detail how such persistence is likely to emerge in the form

of a “daisy chain” or “round robin.” A fail can be the result of operational disruptions such

as the September 11 attacks, but is mostly determined by the incentive embedded in the cost

of failing. In deciding whether to reverse repo in the collateral necessary to avoid failing, a

market participant must weigh the cost of borrowing the security with the cost of failing to

settle. In practice, the borrowing cost (or reverse repo rate) can exceed the cost of failing to

deliver. These events are most often associated with strong demand due to short positions

or limited security supply. Furthermore, although the likelihood of failing to settle does not

appear to be correlated with changes in monetary policy, there is strong evidence it is related

to the general level of interest rates. For example, in contrast to the weeks after September

11th, in 2003 when the FOMC lowered the federal funds rate to 1 percent, we can see in

Figure 6 a corresponding increase in settlement fails for all securities. Such findings are

consistent with the arguments of Fleming and Garbade (2002) who stress that the incentives

to avert a fail are low when interest rates are low. Interestingly, however, we shown that the

timing pattern of fails by security is different. Liquidity pressures appear first in MBS, then

Treasury, and lastly agency.

II.4 Adding the Monetary Authority to the Securitized BankingModel

The transmission of monetary policy from repo activity to securitized banking is outlined by

the stylized flow diagram in Figure 7. The starting point in the figure is a group of obligors,

individuals seeking access to funds. Mortgage borrowers, entrepreneurs, and consumers

seeking loans are all examples.

20

Originators can be traditional banks or mortgage brokers. Upon granting credit to oblig-

ors, originators sell the loan obligations to a bankruptcy remote special purpose vehicle

(SPV). The “true sale” of the obligations to a SPV provides originators with funds to ex-

tend additional credit. (Gorton and Souleles, 2007) SPVs pool the loan obligations and

package them into securities which are bought by a dealer.

Dealers can acquire funds to make markets in these securities via repos with money

market mutual funds and securities lenders. Dealers then sell some of the securities they

hold to other market participants such as hedge funds, commercial banks, and other dealers.

In addition, dealers also extend credit to finance these purchases to other market participants

through reverse repos. Proceeds from outright sales in addition to offsetting reverse repo

agreements provide the dealers with the necessary income to settle their originating repo

obligations thereby keeping the securitized banking model running efficiently.

The final agent in Figure 7 is the monetary authority. For our purposes, we are princi-

pally interested in the trades that occur between the monetary authority (the open market

desk) and the primary dealers. As described previously, these could be either permanent or

temporary open market operations. As an example, in a temporary open market operation,

the open market desk may lend funds to a dealer.

That is, the interaction between a dealer and the open market desk can be the same as a

repo between a dealer and a money market mutual fund. In turn, this affects the amount of

activity that a dealer can undertake with other market participants, including activities with

the SPV. As a result, monetary policy transmits to repo and securitized banking activity.

III Empirical Methodology

In the following section we outline the empirical model we use to identify the transmission

of monetary policy to repo market activity. We begin by defining and summarizing simple

descriptive statistics of the variables used in our study. We then present preliminary evidence

suggesting a link between traditional monetary policy tools and repo market activity. Lastly,

21

we define our empirical model of the transmission mechanism including the assumptions

necessary for identification.

III.1 Data Description

Detailed descriptions of our data and its sources are listed in the appendix. Our interest

is in the response of repo market behavior with respect to changes in monetary policy

instruments including traditional cost of credit measures and System Open Market Account

holdings. The SOMA holdings, effective federal funds rate, and federal funds target rate are

each collected from the Board of Governors H.4.1, H.15, and press releases respectively. Our

repo data is published by the FRBNY and collected from the FR 2004C report.

All of the repo data we use is based upon the report’s July 2001 revision. As a result,

our sample begins July 4, 2001 and ends January 31, 2007, a week before the first bank was

placed into FDIC receivership.29 It is common in the monetary transmission literature to

include a real activity measure and a measure of overall prices. These are included to control

for possible endogeneity due to economic activity, however, most studies are conducted at

lower frequencies than ours.30 Notably, energy prices have been shown in a number of papers

to be an important real activity measure.31 We follow the literature by using the spot price

of West Texas Intermediate (WTI) Crude Oil (Cushing, Oklahoma) as a measure of energy

prices. The price of WTI crude oil is released by the U.S. Department of Energy in the Energy

Information Administration Petroleum Status Report. We calculate a weekly measure of oil

prices by taking an un-weighted average of the daily closing spot price over the specified

time period.

In addition, we include the four week average of initial jobless claims published in the

29The last bank to be placed in FDIC receivership was on June 24, 2005. In addition, Gorton et al. (2013)presents empirical evidence of a structural break beginning one week before our sample end date in the priceof subprime credit risk.

30Early work by Geweke and Runkle (1995) finds that time aggregation does not have a significant impacton inference in studying the effects of monetary policy in standard VAR’s.

More recently, Ghysels et al. (2012) and Chiu et al. (2012) have begun developing VAR methodologiesthat use daily interest rates and monthly measures of real activity.

31For example see Leduc and Sill 2004, Hamilton and Herrera 2004, and Hamilton 2009.

22

U.S. Department of Labor Unemployment Insurance Weekly Claims Report as a measure of

labor market conditions. It is published Thursday for the week ending Saturday before the

release and is revised. New unemployment claims are compiled weekly and show the number

of individuals filing for unemployment insurance for the first time.32

Table 2 presents descriptive statistics of our variables over our sample. Except for interest

rates and the SOMA fraction of Treasury bills to Treasury securities, we transformed all of

our variables into natural logs. The last column lists the coefficient from a regression, in

levels, of the row variable on its one period lag. Other than SOMA repos and deviations in

the federal funds rate from the target rate, our variables display high levels of persistence.

As a result we take first differences and transform our log variables into growth rates.

III.2 Preliminary Evidence

In this section we present results from univariate regressions and an augmented distributed

lag (ADL) model which shows a strong relationship between traditional cost of credit policy

instruments and shadow banking activity. The results introduced here motivate the use

of a VAR model for representation of the monetary transmission mechanism to gross repo

activity.

Net Borrowing by Dealers and Interest Rate Indicators of Monetary Policy

Though our ultimate aim is to study the transmission of monetary policy to repo markets

using a VAR, we begin with a simple single equation analysis. Notably, if monetary policy

transmits to shadow banking activity then we should see a strong correlation between FOMC-

controlled measures of the cost of credit and repo market activity. We begin by looking at

the relationship between the cost of credit in the money market and net repo flows. Table 3

32The literature on real-time forecasts of economic activity finds marginal information in the weekly joblessclaims report to be statistically significant in quarterly forecasts of real GDP. For measures observed dailyhowever, the informational gains are limited (see Gavin and Kliesen (2002), Giannone et al. (2008), andAruoba et al. (2009)).

23

presents a set of simple univariate OLS regressions corresponding to the following levels-on-

levels specification:

yw = α + βxw + uw (1.1)

where yw is our net repo and xw is the our cost of credit measure. The coefficient

β measures the correlation between the cost of credit and net repo flows. As previously

mentioned, we find strong persistence in our variables so we report Newey-West (1987)

standard errors which are robust to serial correlation up to 13 lags. Furthermore, because

the repo data is not seasonally adjusted we report results from regressions that include

weekly fixed effects.

We find robust correlations between our traditional cost of credit measures and net repo

activity. Results for both the target federal funds rate and the effective rate indicate that

higher levels of the interest rate are positively related to net repo activity. Thus, higher

costs of credit in the interbank market are positively related to the level of repo borrowing

among the dealers.

It is also instructive to look at the relationship between indicators of monetary policy

and activity across the maturity structure. First, there is a strong positive relationship

for net repo flows and net overnight and continuing repo funding. Net term flows, while

highly significant, have a negative correlation. In addition to different qualitative responses

to the federal funds rate, there are also significant quantitative differences by maturity. In

particular, the response of net overnight and continuing repo is ten times higher in absolute

terms than term activity. Consistent with this pattern, as observed by the different R-

squared numbers across regressions, monetary policy appears to be a much more significant

factor in overnight and continuing repo activity than term.

These results describe a richer relationship between shadow banking activity and mon-

etary policy than previously identified in the literature. Not only is the level of the federal

funds rate highly correlated with the amount of net repo financing conducted by primary

24

dealers, the correlation extends to the maturity structure. However, when thinking about

monetary transmission mechanisms, the interest rate channel suggests we should find a neg-

ative correlation between interest rate policy measures and net repo. On the other hand,

Woodford (2010) observes that the rate for federal funds may not fully capture conditions

in the money market. As a result, we also look at the correlation between deviations of

the funds rate from the target rate and net repo behavior. Towards that end, we define

the variable ‘MISS ’ as the weekly average target rate subtracted from the weekly average

effective rate in the funds market. For example, if the MISS is positive, this implies that the

federal funds market is trading more tightly than desired by the FOMC. Notably, we find

that such imbalances are significant and negatively related to net repo and net overnight

and continuing repo activity. As such, the coefficient on the MISS suggests that while the

level of reserves is not related to large amounts of net funding in the repo market, net repo

funding is responsive to the interest rate channel.

In terms of net financing, the coefficient is still positive and significant. However, as

we would expect from Figure 5, the relationship is muted. Again, we find the qualitative

response of net overnight and continuing and net term financing varies. Nevertheless, the

magnitudes are much closer than the estimates for net repo. Interestingly, the MISS is not

a significant predictor of either net financing flows or net overnight financing. We do find a

marginally significant negative relationship with respect to net term activity.

Gross Flows

In the previous section we presented regressions that showed net flows into the repo market

are highly correlated with the federal funds rate and reserve imbalances. However, while

the level of interest rates is informative, it is unable to indicate how repo activity responds

to changes in monetary policy. Furthermore, fragility in the shadow banking system can

emerge as a particular piece of collateral is used to finance multiple arrangements with

multiple counterparties. As a result, we now turn to studying how changes in the cost of

25

credit affect gross repo activity. In particular, the gross repo flows in Figure 5 suggest the

series have non-stationary time series properties. Consequently, a simple static specification

such as (1) from the previous section is inappropriate. Instead, we model gross repo activity

through an ADL(13,13) model as follows:33

∆yw = α + β∆xw + Σ13i=1γi∆xw−i + Σ13

i=1θi∆yw−i + uw (1.2)

In equation (2), we look at how current and past changes in monetary policy affect gross

repo growth. Table 5 presents the results for the impact multipliers associated with different

policy indicators. In contrast to the results for net flows, the effective federal funds rate is

significantly related to gross activity but not the target. However, the difference is likely to

be due to the increase in the number of parameters to estimate in the ADL specification

in (1.2). The MISS is also significant. As in the case of the net flows, there is evidence

of maturity substitution from term to overnight contracting arrangements. However, the

magnitude for the multiplier is larger in absolute terms for term activity than overnight

activity. As this indicates that higher rates drive gross flows down, the results aggregated

across maturities are consistent with the lower volume of gross activity.

III.3 Modeling the Open Market Desk

The preceding single-equation analysis provides a number of motivating issues. For example,

there is an array of evidence that seems to indicate that policy-tightening is associated

with both changes in levels of activity and the maturity structure of repo arrangements.

However, there is one major concern that limits our insights from Section 3.2. That is, it

is hard to argue there is a clear causal mechanism from policy to repo activity. One well

known solution is to estimate a VAR model with a limited set of variables. In conjunction

with a suitable identification scheme, a VAR model allows one to isolate the endogenous

response of the central bank from purely exogenous variation. However, because we are

33Following Adrian and Shin (2010) we choose a lag length of 13 weeks.

26

using weekly observations, any unexplained variation in our chosen indicator of the stance of

Federal Reserve policy likely reflects the actions of the open market desk designed to promote

federal funds rate targeting. We therefore modify the standard estimation procedure slightly

in keeping with institutional realities. Specifically we identify a monetary policy shock with

the residuals from the following regression equation:

pw = Ψ(ΩDESKw

)+ λpεp

w (1.3)

In other words, the chosen policy instrument (pw) is equal to a linear combination of the

current economic state observed by the open market desk when setting the policy instrument(ΩDESKw

)each week and a positive serially uncorrelated shock (λpεp

w) orthogonal to the desk’s

observed state of the economy as a result of information lags. Using the model’s reduced form

representation, we measure the dynamic responses of repo market activity to our monetary

policy shock by estimating the following VAR:

Ψw = A(L)Ψw−1 + uw (1.4)

where A(L) ≡ I −A1L is defined as the auto-regressive lag polynomial of order one, and

uw is a vector of reduced form residuals.34 After rearranging the VAR, we equate uw to the

structural economic shocks εw as:

uw = Λεw (1.5)

Identification of the underlying structural monetary policy shock, εpw, requires a set of

restrictions to be imposed upon Λ. We elaborate on our identification scheme in the following

sub-section.

34For simplicity, we generalize by considering a first order VAR and excluding deterministic regressors.

27

III.4 Identification

Our aim is to present evidence on the transmission of monetary policy to repo market

activity. However, in addition to including a measure of policy and repo activity in Ψw, our

identification assumptions also require the inclusion of a limited number of variables which

adequately capture the state of the real economy. For obvious reasons we would like to use

real GDP and the GDP deflator to measure macroeconomic productivity and the general

price level respectively. However, we are forced to use new indicators of broad macroeconomic

conditions because the measures found in the literature are unavailable week to week.35 We

therefore use the four week moving average of initial jobless claims and the spot price of

oil as our measures of economic activity and energy prices respectively. Figure 8 provides

supporting evidence in that these two high frequency measures closely approximate their low

frequency analogues. In addition, we begin our analysis by assuming the effective federal

funds rate is the relevant monetary instrument and gross repo activity by the primary dealers

is a measure of shadow banking activity. To summarize, our VAR includes the following four

variables:

ΨTw = [DL CLAIMSw, D FFw, DL OILw, DL Rw] (1.6)

As a result, the relationship between our reduced form residuals and structural distur-

bances defined by equation (5) can now be expressed as:

u1w

u2w

u3w

u4w

=

b11 b12 b13 b14

b21 b22 b23 b24

b31 b32 b33 b34

b41 b42 b43 b44

εDL CLAIMSw

εD FFw

εDL OILw

εDL Rw

(1.7)

As denoted in the previous section, each structural disturbance is serially uncorrelated

and has a covariance matrix equal to the identity matrix. If we replace E[uwu

Tw

]= Σu

by its sample analogue, Σu has n(n+1)2

= 10 free parameters and the Λ matrix contains

35Bernanke and Mihov (1998) construct a bi-weekly measure of industrial production and the CPI usinginterpolation.

28

16 elements. Therefore, n(n−1)2

= 6 additional restrictions are necessary and sufficient to

estimate an exactly identified system.

Our main identification assumption, proposed by Bernanke and Blinder (1992), is that

real activity responds to changes in monetary policy with a lag. However, we argue that

repo volumes and oil prices respond contemporaneously to monetary shocks. We therefore

impose a Choleski decomposition such that b1,2 = b1,3 = b1,4 = b2,3 = b2,4 = b3,4 = 0 with the

following recursive structure:

u1w

u2w

u3w

u4w

=

b11 0 0 0b21 b22 0 0b31 b32 b33 0b41 b42 b43 b44

εDL CLAIMSw

εD FFw

εDL OILw

εDL Rw

(1.8)

In other words, consistent with our identification assumptions, the restricted Λ matrix

is equivalent to monetary policy affecting financial market activity contemporaneously and

real activity with a lag.

The Information Content of Interest Rate Indicators of Fed Policy

A key issue in using VAR-based models of policy is whether the shocks for policy that are

constructed are purely exogenous. Standard analysis in the monetary policy literature uses

observations at low frequencies such as the monthly or quarterly basis. One solution in the

literature is to use daily observations of policy actions. As a result of these concerns, it is

also questionable whether the effective rate indicates changes in policy rather than changes

in demand at the weekly frequency. One simple way to investigate this issue is to look at the

reduced-form relationships at the daily level. Table 6 presents forecast regressions between

changes in the effective federal funds rate, changes in the target rate, and the MISS variable.

Regressions of the target on lagged changes in the effective rate indicate the effective rate

has limited information content about the target. In contrast to the regression of the target

on the effective rate, the target is an important forecasting variable for the effective rate.

As the effective rate is largely influenced by changes in the target but the target does not

29

appear to respond to the effective rate, this indicates that the effective rate is an important

indicator of Federal Reserve policy even at the daily frequency.

Figure 9 provides implicit graphical evidence of the relationship between the MISS, the

target rate, and the effective rate. The tightening and easing cycle during our sample shows

an apparent systematic relationship between the MISS and the target. Table 6 provides

statistical evidence that there is significant information content in the MISS. These results

are consistent with arguments by Demiralp and Jorda (2002) that the MISS is an important

measure of expected changes in Fed policy.

The Weekly Supply and Demand for Bank Reserves

In this sub-section, we pursue an alternative approach to defend our measure of monetary

policy shocks based upon a strategy proposed by Bernanke and Blinder (1992). In order to

claim that the effective federal funds rate is an indicator of monetary policy at the weekly

frequency, Bernanke and Blinder begin by setting up a simple 3-variable VAR with the

federal funds rate, nonborrowed reserves, and required reserves. They then run a regression

of the innovations of the funds rate on innovations to nonborrowed reserves. Their null

hypothesis is that if the federal funds rate is unresponsive to nonborrowed reserves, then the

innovations are representative of supply shocks in the money market rather than demand

shocks. However, their sample period is different than ours.

Therefore, we find it necessary to run the same scheme during our sample. We begin

by presenting a simple partial-correlation matrix among the three variables found in Table

7. The results are supportive of an interest rate targeting regime. As the federal funds rate

is uncorrelated with the innovations to money demand, the innovations to the funds rate

are measures of supply shocks. Moreover, required reserves are correlated with nonborrowed

reserves but not the federal funds innovations. Thus, required reserves can be an instrument

for nonborrowed reserves. In addition to the lack of correlation of the federal funds rate with

the two measures of money demand in the correlation matrix, the t-stat in the regression of

30

funds innovations on nonborrowed reserves in Table 9 is small (0.009) along with its economic

significance.

The Open Market Desk’s Reaction Function

While we have already shown evidence that innovations to the funds rate are not correlated

with transactions demand for money, there may still be questions about our identification

methodology at the weekly frequency.

For example, there may be questions about the recursive ordering. In particular, we as-

sume that financial variables can react contemporaneously to the funds rate but real activity

does not. Hence, our measures of shocks would be inaccurate due to misspecification of the

identification assumptions. As a way of establishing the validity of our results, we begin by

studying the reaction of the federal funds rate in response to unemployment shocks (through

shocks to the four-week moving average of claims) and price level shocks (through the spot

price of West Texas intermediate crude). That is, we are attempting to establish that the

results from our exercise are consistent with plausible desk reaction functions.

First, we identify shocks to the labor market. Figure 10 shows a shock to claims is

positive, significant and peaks in about week 3. In turn, the federal funds rate declines and

is statistically significant for up to three weeks after the shock to claims. Thus, it appears

that the desk reacts in a reasonable way to adverse labor market shocks. Second, shocks to

energy prices push the federal funds rate in the opposite direction. Thus, our estimates for

the desk reaction function behave the same way as those found by Bernanke and Blinder

(1992).

IV Results

The first section focuses on the response of the effective federal funds rate, jobless claims,

and the spot price of oil after a shock to the federal funds rate. The second section focuses

31

on the behavior of primary dealers’ repo activity after a shock to the federal funds rate and

compares this behavior with the responses of dealer repo activity by contract maturity.

IV.1 Benchmark Response – An Increase in the Cost of Credit

The benchmark specifications for the VARs include one quarter of lagged variables, a con-

stant, and weekly fixed effects since the FR 2004 report is not adjusted for seasonality. We

begin by studying the impulse responses for the recursive VAR ordered: DL CLAIMS, D FF,

DL OIL, DL R. We report 90% probability intervals for impulse responses. (Sims and Zha,

1999)

The Federal Funds Rate and Real Activity

The response of the federal funds rate to its own shock is plotted in Figure 12. The federal

funds rate responds by increasing contemporaneously approximately 6 basis points and is

significantly different from zero after one year. However, looking at Figure 13, neither

initial jobless claims or the spot price of oil react much to federal funds rate shocks. There

are a number of reasons why this is viable in our sample. First, it might be due to the

high frequency nature of our data. Second, it could also be due to the sample period that

we study. In particular, many observers have pointed out that monetary policy was not

particularly effective around the time of the recession in 2001 and the latter jobless recovery.

Notably, the FOMC continued to lower the federal funds rate to the historically low 1% rate

in June 2003 where it remained for one year. Third, it could also be the result of modest

policy adjustments implemented during the “Great Moderation.” For example, Angrist et al.

(2013) find it takes around 18 months for industrial production and employment to respond

to policy shocks in studying the U.S. economy from 1989 - 2010. Moreover, there is no

inflation response out to two years.36

36Caplin and Leahy (1996) argue that modest changes in policy are likely to be ineffective in stimulatingmacroeconomic activity.

32

The Level and Maturity Structure of Repo Financing

We begin by studying the impact of the cost of credit shocks on the total amount of repo ac-

tivity presented in Figure 14. The contemporaneous response of total repurchase agreements

is negative and statistically significant. The point estimate indicates that the contemporane-

ous 6 basis point increase in the effective federal funds rate is associated with a decline in the

level of total repo financing by nearly 0.4%. By comparison, the single equation analysis of

Adrian and Shin (2009) found that an increase in the federal funds rate of 1% (or 100 basis

points) would be associated with about a 4% decline in repo activity. Instead, our empirical

model suggests that the same increase in the effective rate would be associated with nearly

an 8% decline. Thus, our results indicate that policy shocks would have a larger impact

on repo volumes than previously found. We view that our VAR analysis with additional

controls for endogeneity is more suggestive of an exogenous policy shock. Consequently, the

impact of policy appears to be stronger.

The advantage of our VAR approach is the ability to study the cumulative effects of

shocks over a period of time. In particular, we find that the peak (negative) response occurs

around the fifth week after a shock to the effective rate. While the federal funds rate is

virtually the same as the level following the initial shock, the impact on total repo volumes

is even lower at around a 7% decline. Again, this highlights that our evidence indicates that

the effect of monetary policy on repo markets is greater than previous work. Though the

point estimate is not statistically significant, it continues to be negative one year after the

shock.

In contrast to the impact on total repo activity, how do cost of credit shocks affect the

maturity structure of repo financing? Notably, many observations from the crisis suggest that

adverse money market conditions contributed towards a shift towards short-term financing.

For example, many have suggested that the failure of Lehman Brothers severely affected

repo markets to the point where only overnight financing was available.

Moreover, policymakers still debate the concerns that they have about rollover risk among

33

institutions with significant maturity mismatch. Notably, Figure 5 demonstrates that pri-

mary dealers’ net borrowings are short-term while they engage in relatively long-term lending

in the repo market. As another example of maturity mismatch, Adrian and Fleming (2005)

cite the delay that dealers have in placing their inventories of mortgage-backed securities.

In particular, they observe that beginning in 2001 net financing among primary dealers was

significantly larger than their net positions in mortgage backed securities. One possible ex-

planation is that the gap increased as dealers were holding larger inventories of mortgage

obligations that would settle long-term but were financed on a daily basis. (Please refer to

Figure 7 for additional discussion – one of the final steps in the intermediation chain of se-

curitized banking takes place when dealers eventually place securities through sales to other

intermediaries such as hedge funds and commercial banks.)

Figure 14 also plots the impulse response of maturities of repos in response to the cost

of credit shocks. In response to the shock to the target rate, term volumes are negative and

statistically significant beginning in week 1. The point estimate for the decline is about 1%

lower in response to about a 5 basis point shock. As the standard policy change is 25 basis

points, our estimates imply that a standard rate hike would correspond to a 5% decline in

term repo volumes. There is weak evidence of maturity substitution – as the point estimate

for term activity is negative and significant, the point estimate for open activity is positive

but generally insignificant.

V Discussion

V.1 Alternate Interest Rate Measures of Monetary Policy

In addition to shocks to the effective rate, we study impulse response functions from shocks

to the main policy tool - the target for the federal funds rate. The point estimate for the

contemporaneous response is negative and shows significance from the first week after the

shock for a total of three months. Moreover, the elasticity of total repo activity in response

34

to shocks to the target is the same as shocks to the effective rate. As the behavior of repos

following shocks to the target so closely mirrors shocks to the effective rate, we argue that

shocks to the effective rate represent supply shocks to money markets in the same manner

as shocks to the target. Though the differences are minor, the impulse response functions

for the target are somewhat more smooth than shocks to the effective rate. This potentially

reflects that information in the target incorporates lower frequency broad macroeconomic

conditions in comparison to weekly movements in unemployment claims and oil prices.

Our final type of cost of credit shock is the deviation of the effective rate from the target

rate, the MISS. As discussed by Demiralp and Jorda (2002), the deviation of the funds rate

serves as an indicator of reserve imbalances in the money market. In comparison to shocks

to the effective rate and the target, an initial shock does not signal the beginning of a cycle

over time. Thus, one could interpret that such shocks are not long-lasting. We find that

these deviations are associated with lower volumes of repo activity. However, such losses

are concentrated in overnight repos. Moreover, overnight repos secured by MBS contract

the most. (See the third row and column of Figure 17.) This likely reflects that distortions

in reserves have the greatest impact on activity in the most interest-sensitive sectors of the

economy.

V.2 The Response of Aggregate Financing

As previously discussed, repurchase agreements are a subset of the overall collateralized bor-

rowing the primary dealers report. In this section we compare the responses of gross financing

arrangements with gross repo. Figure 15 plots the responses of our three financing measures

to our three cost of credit shocks. The first, second, and third columns plot responses to

a shock in the effective federal funds rate, target, and MISS respectively. Qualitatively,

all of the responses we find are very similar to the ones found for repo activity. Changes

in the level of the federal funds rate lead to declines in gross financing and financing with

longer maturities. However, overnight financing activity increases with a delay. An increase

35

in reserve imbalances leads to a delayed decline in gross financing and financing overnight.

However, there are a few subtle differences. Looking at the first row, we see the decline in

financing activity in response to an increase in the federal funds rate has a similar magnitude

to the response of gross repo but in contrast, greater persistence. In particular, the response

of gross financing is significant for over 6 months in contrast to a month for repo activity. In

addition, we find the response of overnight financing activity to shocks in the federal funds

rate and the target both show similar point estimates but weaker significance.

V.3 The Response of Financing Fails

In addition to affecting the level of repo activity, one obvious question involves understanding

whether unanticipated monetary policy actions contribute to instability in the repo market.

One sign of instability would be if there is a change in the level of fails in the system in

response to monetary policy shocks, indicating that agreements do not settle promptly. The

first row of Figure 16 plots the responses of financing fails to our cost of credit shocks. There

does not appear to be a direct relationship between monetary policy and fails in repos.

However, reserve imbalances (as indicated by the MISS ) affect the level of fails.

V.4 The Response of Security Asset Classes

We are also interested in the securities that are used for collateral across repurchase agree-

ments. Figures 6 plots net financing by security class and by maturity. As previously

discussed, dealers activity engage in maturity mismatch where they borrow short-term and

lend long-term. Figure 5 demonstrates that the majority of net repo activity occurring

overnight takes place through pledging MBS as collateral. By comparison, Figure 6 shows

the types of collateral accepted in return for lending funds.

We begin by looking at securities that are pledged in financing transactions by the dealers.

Financing transactions can take place in two forms: (a) securities pledged for cash and (b)

securities pledged in order to borrow other types of securities. Standard repo transactions

36

(in which securities are pledged for cash) represent the majority of transactions in which

securities are pledged. Consistent with our finding that cost of credit shocks drive down

total repo activity, we observe in Figure 17 a decrease in the use all of types of securities as

a form of collateral.

Yet, we view that our results tell interesting story about a “collateral cycle” in financing

behavior after a shock to the effective rate. First, the initial stages of a shock to the effective

rate generally are associated with a tightening cycle in which rates continue to rise over

time. In the initial stages of the cycle, rates do not rise by much. Consequently, the

first column of Figure 18 shows there is some substitution of collateral towards agency

and corporates in overnight and continuing agreements. Over time, however, rates increase

further such that interest rate sectors of the economy are under pressure. After about

twelve weeks, less transactions take place through agency and corporates while treasuries

are positive and statistically significant. The effects for MBS are insignificant as there may

be two conflicting factors. First, there is weak evidence of maturity substitution from term

financing to overnight – the bulk of overnight net financing occurs through MBS. However,

the housing sector is very interest-sensitive which would lead to substitution out of MBS as

a form of collateral.

One might conclude there is indirect evidence of policy affecting the stability of the

repo market through liquidity disruptions. Shocks from deviations of the effective rate are

associated with a temporary increase in the number of fails involving treasuries (see Figure

20). That is, in response to imbalances in money markets, there is a substitution towards the

highest quality of collateral. As a result, there is an increase in demand for treasuries. One

interpretation is that money markets would be moving ahead of the central bank whenever

deviations occur. The deviations are eventually followed by a response from the FOMC.

However, in such cases, the central bank is simply following a tightening cycle initiated in

money markets.

37

V.5 Repo Activity and the System Open Market Account

The FOMC influences the money market not only through the cost of credit but also through

changes in the level of bank reserves. The level of reserves is changed through open market

operations which can be temporary and or permanent. In the following subsection we ex-

plore the response of dealer repo activity after a change in the composition of the SOMA.

Temporary operations are conducted through repurchase and reverse repurchase agreements

while permanent operations occur through treasury securities.

Temporary Liquidity Injections

The impulse responses for the recursive VAR ordered DL CLAIMS, L FEDR, DL OIL, DL R

after a one standard deviation increase in L FEDR are plotted in Figure 21. As described

previously, temporary repos initiated by the desk are tri-party with the dealers as counter-

parties. Figure 17 illustrates the transmission mechanism from the actions of the desk. A

shock to repos conducted by the desk is positive, significant, peaks in week 20 and remains

significant after one year. It takes some time before L FEDR shocks to affect repo activ-

ity. Following the peak in L FEDR, term repo activity is positive and significant. As the

data for L FEDR are in log levels, the contemporaneous shock represents a shock to repos

initiated by the desk of around $200 million. Cumulatively, the increase peaks around $630

million. In response to the peak increase, term repo activity increases by around 1.5%. At

the beginning of the sample, term repos by the dealers were equal to about $860 billion.

Thus, the $630 million shock from the desk would have translated into an increase of term

repos by nearly $13 billion.37

There are interesting findings from looking at how repo arrangements are collateralized.

In this respect, there are two competing factors. First, an injection of liquidity promotes

repo activity. It does so in two ways. First, injections may lead to longer-term credit

arrangements. Second, they can also affect the types of securities that are used as collateral.

37A similar exercise by Hamilton (1997) studies the impact of sustained shocks to nonborrowerd reserveson the federal funds rate. In particular, he finds that a $30 million shock is associated with an increase of10 basis points.

38

In this manner, the effects L FEDR shocks are hard to pin down a priori. We therefore need

to turn to the impulse response functions to determine the net impact.

As a first step, we begin by focusing on collateralization. That is, the first column

of Figure 22 looks at the response of aggregate financing activity according to how the

arrangements are collateralized. At first, there is a substitution away from safe securities

such as Treasuries and Agency securities. During the first six weeks after the injection, point

estimates are negative and statistically significant. Thereafter, there is a movement towards

securities in the credit-sensitive sectors of the economy. Point estimates for corporates are

positive and significant after the eighth week. The point estimate for MBS as collateral

behaves in a similar way, but it takes longer to reach significance. Presumably, this reflects

that the housing sector is the sector in the economy which is more sensitive to liquidity.

We proceed by dis-aggregating further by studying the maturity structure along with

collateralization. Impulse response functions plotted in Figure 24 indicate that the short-

term response is a movement away from Treasuries as collateral for term arrangements. At

the same time, Figure 23 shows an increase in short-term arrangements that are collateralized

with corporate securities. Presumably, this indicates that there is a movements towards

sectors that are credit sensitive after an injection of liquidity. After a quarter of time, there

is a movement towards term funding and an increase in the use of each form of collateral.

A shock to L FEDR appears to be linked to stability in the repo market. After the shock,

fails decline. In particular, Figure 25 shows a decline in arrangements secured by MBS.

Temporary Liquidity Withdrawals

In contrast to repos initiated by the desk, reverse repos are bilateral, delivery vs. payment

transactions with money market mutual funds. Instead of temporarily injecting liquidity

to money markets, reverse repos temporarily remove liquidity. The significance of money

market funds as counterparties is that these institutions lend heavily to the dealers. Thus,

reverse repos would pull funds away from private institutions. Contemporaneously, the

39

shock to reserve repos (Figure 12) increases by about 6% but later declines to 4%. This may

suggest that the initial implementation of monetary tightening occurs through temporary

open market operations and then is substituted by permanent open market operations. After

the initial response, there is a sustained loss of term repo activity of nearly 2%. There is

strong evidence of maturity substitution after the shock. During the same time frame of the

contraction in term repo activity, open volumes increase by around 3/4 of a percent.

Impulse response functions for securities out mirror the response for term repurchase

agreements and we therefore do not plot them. The impact on treasuries used as collateral

is particularly strong. In fact, Figure 24 shows maturity substitution from term to overnight

explains the increase in overnight and continuing agreements secured by Treasuries. The

second column of Figure 25 shows a short-term negative impact on fails in treasuries which

seems to be driven by the increased supply of collateral from reverse repos conducted with

the desk.

Securities Held Outright

The FOMC can also permanently change the level of reserves by selling and buying Treasury

securities in the open market. The first and second columns for Figures 26 plot the response

of our variables of interest to an unexpected increase in the System Open Market Account

holdings of Treasury securities and Treasury bills respectively. Overall these results show

little significance although the responses of dealer repo activity to a shock to the System

Open Market Account holding of Treasury bills shows a significant negative contemporaneous

response and dealer repo activity is negative but insignificant.

Recent monetary actions through the Fed’s balance sheet policies have been geared to-

wards impacting financial market conditions. Consequently, the composition of the System

Open Market Account could have an impact on dealer repo activity. We investigate this

claim by constructing the ratio of Treasury bills to Treasury securities in the SOMA as an

indicator of the SOMA’s composition. Figure 26 column 3 shows the effect of an unexpected

40

1 percentage point increase in dealer repo, overnight, and term activity. While overnight

and continuing activity is negative and significant contemporaneously, only the level of repo

activity seems to show a significant persistent decline. Of all the response functions showing

securities used as collateral in Figures 27 to 29, the effect on agency securities is positive

and highly significant presumably due the fact that there is a lower supply of Treasuries.

Consequently, there is a movements towards the use of Treasuries as there is an increase in

Treasuries held on the Fed’s balance sheet.

In comparison to the effects of permanent open market operations through Treasury se-

curities, we find more evidence of significant activity. Presumably, this reflects that bills

have a central role in money market activity than other securities. Due to the lower volume

of term repo activity in response to the SOMA’s holding of Treasury bills, there is a con-

temporanous decline in agency securities used as collateral in overnight arraingements. As a

result of the maturity substitution towards overnight arrangements, there is also an increase

in the use of agency securities for overnight funding. The withdrawal of collateral from repo

markets as a result of the purchases leads to an increase in fails. The majority of fails are

concentrated in MBS.

Our final look is to study the effects of an increase in the proportion of treasury bills

relative to total treasuries (bills + securities) held in SOMA. Such shocks would represent

the inverse of the Federal Reserve’s recent Maturity Extension Program, also known as

“Operation Twist.”

In this manner, one might interpret as results as providing insight into a “Maturity

Reduction Program.” First, note that a shock to the proportion of bills held is associated

with a lower amount of repo activity over time. Consistent with this observation, we find

there is a short-term decrease in the use of each type of asset as collateral.

The results are somewhat stronger for term agreements than total repo activity. In-

terestingly, we find that shifting the maturity of the SOMA portfolio towards short-term

securities leads to a significant increase in fails. Thus, our results suggest that the recent

41

MEP likely promoted stability in repo markets. An increase in fails is observed in virtually

every category of security.

V.6 Choosing the Right Policy Instrument: Evidence from Fore-cast Error Variance Decompositions

We conclude our statistical analysis with variance decompositions that show the overall

influence of various policy instruments available to the open market desk on each aspect

of repo activity. Interestingly, Federal Reserve reverse-repos have the largest impact of any

instrument on total repo activity. (See Panel A of Table 10) However, for most time horizons,

the target for the federal funds rate has the largest impact on short-term repo activity. In

contrast, term activity is dominated by reverse-repos.

It is important to note that other methods of intervention by the Desk and money market

indicators also play a role. For example, reserve imbalances as captured by the MISS, are

an important factor in total repo activity. The second largest contributor to overnight repos

is the amount of Fed repos. For term activity, it depends on the horizon. At short horizons,

the MISS is the second largest factor while at longer horizons the target dominates. The

relationships are more clear in looking at the determinants of financing.

In terms of thinking about stability of the repo market, we look at the most important

factors in explaining the level of financing fails. At short horizons, fails are dominated by

the effective federal funds rate. The second most important factor is the amount of reverse-

repos. At longer horizons, the relationships change. The amount of reverse-repos is the

most important factor followed by the MISS. In this manner, our results echo the warnings

of Bernanke and Blinder (1988) who argue that different tools have different impacts.

VI Conclusion

The monetary transmission mechanism of the Federal Reserve System is linked through the

aggregate balance sheets of three counterparties: depository institutions, Federal Reserve

42

Banks, and the primary dealers. Traditionally, macroeconomists have focused on the first

two and ignored the last. However, the financial crisis put more emphasis on understanding

the systemic importance of market based intermediaries such as the primary dealers.

The purpose of our work is to identify whether monetary policy transmits to the shadow

banking system via primary dealers’ gross repo activity. Using a vector autoregression model,

we characterize the monetary transmission mechanism at the weekly frequency and measure

its’ impact on repo market volumes. We find evidence that this is the case. When the open

market desk conducts monetary policy, the structure of the transmission mechanism influ-

ences credit activity in the repo market through the FRBNY’s counterparty relationships.

Furthermore, it has been argued that the class of security in repos can trigger runs. As

instability spirals through the shadow banking system, the commercial banking system is

also at risk. Our results show that FOMC decisions can influence supply and demand for

money substitutes such as repurchase agreements and may lead to unintended consequences.

Moreover, the FRBNY’s relationship with the primary dealers, is that of counterparty, not

regulator. The limited degree of oversight suggests that macroprudential policy should be

an important concern for monetary authorities.

43

References

Adrian, Tobias, Brian Begalle, Adam Copeland, and Antoine Martin. 2013. Repo and secu-rities lending. In Quantifying Systemic Risk Measurement: NBER Research ConferenceReport Series, ed. J.G.Haubrich, A.W. Lo. University of Chicago Press. Forthcoming.

Adrian, Tobias, Michael J. Fleming. 2005. What financing data reveal about dealer leverage.Current Issues in Economics and Finance, 11(3): 1-7.

Adrian, Tobias, and Hyun Song Shin. 2010. Financial intermediaries and monetary eco-nomics. In Handbook of Monetary Economics, ed. Benjamin M. Friedman and MichaelWoodford, 1(3), Chapter 12: 601-650, Elsevier.

——-. 2009. Money, liquidity, and monetary policy. American Economic Review Papers &Proceedings, 99(2): 600-605.

——-. 2008. Financial intermediaries, financial stability, and monetary policy. Federal Re-serve Bank of New York Staff Reports, no. 346.

Andreou, Elena, Eric Ghysels, and Andros Kourtellos. Should macroeconomic forecastersuse daily financial data and how? Forthcoming Journal of Business & Economic Statistics.

Angrist, Joshua D., Oscar Jorda, and Guido Kuersteiner. 2013. Semiparametric estimatesof monetary policy effects: string theory revisited. NBER Working Paper, No. 19355.

Arouba, S. Borag and, Francis X. Deibold, and Chiara Scotti. 2009. Real-time measurementof business conditions. Journal of Business & Economic Statistics, 27(4): 417-427.

Bech, Morten, Elizabeth Klee, and Victor Stebunovs. 2011. Arbitrage, liquidity and exit:the repo and federal funds markets before, during, and emerging from the financial crisis.Board of Governors Finance and Economics Discussion Series, 2012-21: 1-54.

Begalle, Brian, Antoine Martin, James McAndrews, and Susan McLaughlin. 2013. The riskof fire sales in the tri-party repo market. Federal Reserve Bank of New York Staff Reports,No. 616: 1-38.

44

Bartolini, Leonardo, Spence Hilton, Suresh Sundaresan, Christopher Tonetti. 2010. Col-lateral values by asset class: evidence from primary securities dealers. The Review ofFinancial Studies, 24(1): 248-278.

Bernanke, Ben S.. 2012. Fostering financial stability. 9 April.

——-. 2010. Statement before the financial crisis inquiry commission. 2 September

——-. 2007. The financial accelerator and the credit channel. Speech at The Credit Channelof Monetary Policy in the Twenty-first Century Conference. 15 June.

Bernanke, Ben S., and Alan S. Blinder. 1992. The federal funds rate and the channels ofmonetary transmission. American Economic Review, 82(4) (Sep.): 901-921.

——-. 1988. Credit, money, and aggregate demand. American Economic Review. 78(2)(May): 435-439.

Bernanke, Ben S., and Ilian Mihov. 1998. Measuring monetary policy. Quarterly Journal ofEconomics, 113(3): 869-902.

Board of Governors of the Federal Reserve System. 2011. Federal Reserve issues FOMCstatement. Press Release, 20 June.

Brunnermeier, Markus K.. 2009. Deciphering the liquidity and credit crunch 2007-2009.Journal of Economic Perspectives, 23(1) (Winter): 77-100.

Caplin, A. and J. Leahy. 1996. Monetary Policy as a Process of Search. American EconomicReview, 86(4): 689-702.

Chiu, Ching Wai, Bjørn Eraker, Andrew T. Foerster, Tae Bong Kim, and Hernan D. Seoane.2012. Estimating VAR’s sampled at mixed or irregular spaced frequencies: a Bayesianapproach. Federal Reserve Bank of Kansas City Research Working Paper, RWP 11-11.

Copeland, Adam, Antoine Martin, and Michael Walker. 2012a. Repo runs: evidence fromthe tri-party repo market. Federal Reserve Bank of New York Staff Reports, No. 506.

Copeland, Adam, Darrell Duffie, Antoine Martin, and Susan McLaughlin. 2012b. Policyissues in the design of tri-party repo markets. Unpublished manuscript, federal ReserveBank of New York.

Demiralp, Selva, and Oscar Jorda. 2002. The announcement effect: evidence from openmarket desk data. FRBNY Economic Policy Review, (May): 29-48.

Dudley, William C.. 2013. Reflections on the economic outlook and implications for mone-tary policy. 23 September.

45

Duffie, Darrell. 2010. The failure mechanics of dealer banks. Journal of Economic Perspec-tives 23(1) (Winter): 51-72.

Federal Reserve Bank of New York. 2013. Quarterly trends for consolidated U.S. bankingorganizations. Federal Reserve Bank of New York Research and Statistics Group, SecondQuarter 2013.

——-. 2010. Tri-party repo infrastructure reform white paper. Federal Reserve Bank of NewYork Markets Group, May.

——-. 2002. Domestic open market operations during 2001. Federal Reserve Bank of NewYork Markets Group, February.

Financial Crisis Inquiry Commission. 2011. The financial crisis inquiry report. New York:Public Affairs.

Fleming, Michael J., and Kenneth D. Garbade. 2005. Explaining settlement fails. CurrentIssues in Economics and Finance, 11(9).

Garbade, Kenneth D., F.M. Keane, L. Logan, A. Stokes and J. Wolgemuth. 2010. TheIntroduction of the TMPG Fails Charge for U.S. Treasury Securities. Federal ReserveBank of New York Economic Policy Review, (October): 45-71.

Gavin, William T., and Kevin L. Kliesen. 2002. Unemployment insurance claims and eco-nomic activity. Federal Reserve Bank of St. Louis Review, (May): 15-28.

Geithner, Timothy F.. 2008. Reducing system risk in a dynamic financial system. Speechbefore the Economic Club of New York, 9 June.

Geweke, John F., and David E. Runkle. 1995. A fine time for monetary policy? FederalReserve Bank of Minneapolis Quarterly Review, (Winter): 18-31.

Giannone, Domenico, Laucrezia Reichlin, and David Small. 2008. Nowcasting: the real-time informational content of macroeconomic data. Journal of Monetary Economics,55(4): 665-676.

Gorton, Gary, and Andrew Metrick. 2012. Securitized banking and the run on repo. Journalof Financial Economics, 104(3): 425-451.

Gorton, Gary, Andrew Metrick, and Lei Xie. 2012. The flight from maturity. Yale Schoolof Management Working Paper.

Gorton, Gary, and Nicholas S. Souleles. 2007. Special purpose vehicles and securitization.In The Risks of Financial Institutions, University of Chicago Press.

46

Hamilton, James D.. 2009. Causes and consequences of the oil shock of 2007. NationalBureau of Economic Research Working Paper Series, No. 15002.

——-. 1997. Measuring the liquidity effect. American Economic Review, 87(1): 87-97.

Hamilton, James D., and Ana Maria Herrera. 2004. Oil Shocks and aggregate macroeco-nomic behavior: the role of monetary policy: comment. Journal of Money, Credit andBanking, 36(2): 265-286.

King, Matt. 2008. Are the brokers broken? Citi European Quantitative Credit Strategy andAnalysis, 5 September.

Kohn, Donald L., 2008. Monetary policy and asset prices revisited. 19 November.

Krishnamurthy, Arvind, Stefan Nagel, and Dmitry Orlov. 2014. Sizing up repo. Forthcom-ing Journal of Finance.

Krishnamurthy, Arvind, and Stefan Nagel. 2013. Interpreting repo statistics in the flow offunds accounts. NBER Working Paper Series, No. 19389.

Leduc, Sylvain, and Keith Sill. 2004. A quantitative analysis of oil-price shocks, systematicmonetary policy, and economic downturns. Journal of Monetary Economics, 51(4): 781-808.

Martin, Antoine, David Skeie, and Ernst-Ludwig von Thadden. 2013. Repo runs. Forth-coming, Review of Financial Studies

Mian, Atif and Amir Sufi. 2009. The consequences of mortgage credit expansion: evidencefrom the u.s. mortgage default crisis. Quarterly Journal of Economics, 124(4): 1449-1496.

Mills, David C. Jr., and Robert R. Reed. 2008. Default risk and collateral in the absence ofcommitment. Working Paper.

Morris, Stephen, and Hyun Song Shin. 2004. Liquidity black holes. Review of Finance, 8(1):1-18.

Newey, Whitney K., and Kenneth D. West. 1987. A simple, positive and semi-definate,heteroskedastic-ty and autocorrelation consistent covariance matrix. Econometrica, 55(3):703-708. Review of Finance 8(1): 1-18.

Pozsar, Zoltan, Tobias Adrian, Adam Ashcraft, and Haley Boesky. 2010. Shadow banking.Federal Reserve Bank of New York Staff Reports, No. 485.

47

Sims, Christopher A., and Tao Zha. 1999. Error bands for impulse responses. Econometrica,67(5): 1113-1156.

Stock, James H., and Mark W. Watson. 2001. Vector autoregressions. Journal of EconomicPerspectives, 15(4): 101-115.

Stock, James H., and Motohiro Yogo. 2005. Testing for weak instruments in linear IVregressions. In Identification and Inference for Econometric Models: Essays in Honorof Thomas Rothenberg, Donald W.K. Andrews and James H. Stock, 80-108. CambridgeUniversity Press.

Sunderam, Adi. 2012. Money creation and the shadow banking system. Mimeo.

Taleb, Nassim. 1997. Dynamic hedging: managing vanilla and exotic options. New York:John Wiley & Sons.

Task Force on Tri-party Repo Infrastructure Payments Risk Committee. 2012. Final report.15 February.

——-. 2009. Progress report. 22 December.

Velukas, Anton R. 2010. Lehman brothers holdings inc. chapter 11 proceedings examinersreport. United States Bankruptcy Court Southern District of New York, Vol. 3.

Woodford, Michael. 2010. Financial intermediation and macroeconomic analysis. Journalof Economic Perspectives, 24(4): 21-44.

48

Data Appendix

This section provides documentation of the data we use in our empirical analysis. Weekly

financing data is reported on Wednesday by the primary dealers in the FR 2004C Weekly

Report of Dealer Financing and Fails. The report is published by the Federal Reserve Bank

of New York and collects gross outstanding collateralized borrowing and lending including

repo and reverse repurchase agreements in millions of dollars on a gross basis. Security

financing is not seasonally adjusted and include U.S. Treasuries, Federal Agency and GSE

excluding MBS, Federal Agency and GSE MBS, and corporate securities disaggregated by

maturity. Detailed descriptions of the securities that make up these broad categories can be

found in the Reporting Guidelines published by the Board of Governors. The value reported

by dealers is the actual funds paid or received excluding transactions conducted on behalf of

prime brokerage customers. Settlement fails are disaggregated by by security on a cumulative

basis for the reporting period. The number reported is the principle value that was to be

paid of received from cash and financing transaction on theday the trade settled.

Our interest rate variables are collected from two different sources. The weekly effective

federal funds rate is from the Board of Governors H.15 release. It is constructed as the 7 day

average of the daily weighted average rate on brokered trades. The daily target federal funds

rate is published by the Federal Reserve Bank of New York. Average weekly deviations from

the target rate are constructed as the 7 calendar day average of daily federal funds minus

the 7 calendar day average of the target rate.

Weekly holdings of the System Open Market Account is published in the Board of Gover-

49

nors H.4.1 release. On December 13, 2002 the Federal Reserve switched from using Matched

Sale Purchases to Reverse Repurchase Agreements. The two week daily average of nonbor-

rowed and required weekly reserves of depository institutions is from the Board of Governors

H.3 release. The U.S. Energy Information Administration publishes daily spot prices of WTI

Crude Oil in the weekly Petroleum Status Report. The four week moving average of ini-

tial jobless claims is from the U.S. Department of Labor’s Unemployment Insurance Weekly

Claims Report.

50

Table 1.1: The Primary Government Securities Dealers

Dealers at Beginning of Sample Commercial Banking Public Added Withdrawn Name Change MergerABN AMRO Inc. Yes Yes 9/29/1998 12/8/2002BMO Nesbitt Burns Corp. No No 2/15/2000 4/1/2002BNP Paribas Securities Corp. Yes Yes 9/15/2000Banc of America Securities LLC Yes Yes 5/17/1999Banc One Capital Markets, Inc. Yes Yes 4/1/1999 8/2/2004 8/2/2004Barclays Capital Inc. Yes Yes 4/1/1998Bear, Stearns & Co., Inc. No Yes 6/10/1981CIBC World Markets Corp. Yes Yes 5/3/1999Credit Suisse First Boston Corporation No Yes 12/16/1996 1/16/2003– 1/17/2003 1/16/2006Daiwa Securities America Inc. No Yes 12/11/1986Deutsche Banc Alex Brown Inc. Yes Yes 1/12/2001 3/30/2002Dresdner Kleinwort Wasserstein Securities LLC No No 4/30/2001 9/18/2006Fuji Securities Inc. Yes Yes 12/28/1989 4/1/2002Goldman, Sachs & Co. No Yes 12/4/1974Greenwich Capital Markets, Inc. Yes Yes 7/31/1984HSBC Securities, Inc. Yes Yes 6/1/1999 1/17/2006J.P. Morgan Securities Inc. Yes Yes 5/1/2001 8/2/2004Lehman Brothers Inc. No Yes 8/31/1995Merrill Lynch Government Securities Inc. No Yes 5/19/1960Morgan Stanley & Co. Inc. No Yes 2/1/1978Nomura Securities International, Inc. No Yes 12/11/1986Solomon Smith Barney Inc. Yes Yes 9/1/1998 4/7/2003SG Cowen Securities Corporation Yes Yes 7/1/1999 10/31/2001UBS Warburg LLC. Yes Yes 5/1/2000 6/12/2003Zions First National Bank Yes Yes 8/11/1993 3/31/2002Dealers Added During SampleCountrywide Securities Corporation Yes Yes 1/15/2004Cantor Fitzgerald & Co. No No 8/1/2006

Source: Federal Reserve Bank of New York and the Securities Exchange Commission.

Note: The table lists all primary dealers from July 4, 2001 to January 31, 2007. The table includes the name of the dealer, whether it is a commercial banking affiliate, whetherit is publicly traded, and the following primary dealer effective event dates: added, withdrawn, name change, or merger.

51

Table 1.2: Descriptive Statistics: July 4, 2001 to January 31, 2007Obs. Max Min µ σ ρ

Panel A: Net Collateralized BorrowingNet Repo (NETREPO) 292 1304.24 450.10 805.54 230.67 0.994Net Overnight and Continuing Repo (NETREPO OC) 292 1456.44 558.35 927.13 252.21 0.990Net Term Repo (NETREPO TERM) 292 8.26 -232.43 -121.59 39.54 0.788Net Financing (NETFINANCING) 292 369.31 106.81 223.96 49.16 0.832Net Overnight and Continuing Financing(NET OC) 292 864.46 310.80 532.44 125.82 0.962Net Term Financing (NET TERM) 292 -101.83 -562.05 -308.48 99.72 0.964

Panel B: Repurchase AgreementsRepo (DL R) 291 6.72 -25.29 0.25 4.12 0.978Overnight and Continuing Repo (DL OC R) 291 21.31 -20.17 0.32 4.40 0.986Term Repo (DL TERM R) 291 19.43 -57.88 0.15 10.15 0.845

Panel C: Securities OutSecurities Out (DLALL OUT) 291 6.66 -22.40 0.25 3.75 0.982Overnight and Continuing Securities Out (DLALL OUT OC) 291 21.04 -19.76 0.34 3.97 0.986Term Securities Out (DLALL OUT TERM) 291 18.79 -56.81 0.14 9.80 0.867US Securities Out (DLALL OUT US) 291 7.63 -27.89 0.23 5.15 0.975Agency Securities Out (DLALL OUT AGENCY) 291 13.65 -22.25 0.13 3.98 0.899MBS Securities Out (DLALL OUT MBS) 291 14.33 -12.62 0.31 4.67 0.978Corporate Securities Out (DLALL OUT CORP) 291 18.70 -26.47 0.50 3.80 0.992Overnight and Continuing US Securities Out (DLOUT OC US) 291 21.82 -27.33 0.35 5.47 0.984Overnight and Continuing Agency Securities Out (DLOUT OC AGENCY) 291 39.57 -42.34 0.19 5.64 0.885Overnight and Continuing Agency MBS Securities Out (DLOUT OC MBS) 291 26.37 -25.85 0.35 6.37 0.967Overnight and Continuing Corporate Securities Out (DLOUT OC CORP) 291 26.09 -30.65 0.49 4.55 0.988Term US Securities Out (DLOUT TERM US) 291 18.03 -63.65 0.12 11.97 0.852Term Agency Securities Out (DLOUT TERM AGENCY) 291 31.08 -59.18 0.03 8.85 0.868Term Agency MBS Securities Out (DLOUT TERM MBS) 291 38.71 -37.95 0.24 7.98 0.948Term Corporate Securities Out (DLOUT TERM CORP) 291 66.97 -50.71 0.56 9.61 0.986

Panel C: Settlement Financing FailsAll Fails (DLALL FAIL) 291 238.46 -174.34 -0.17 54.59 0.748US Fails (DL FAIL US) 291 342.11 -253.60 -0.49 74.16 0.777Agency Fails (DL FAIL AGENCY) 291 279.38 -186.10 -0.36 50.87 0.819MBS Fails (DL FAIL MBS) 291 265.74 -117.81 -0.04 93.37 0.546Corporate Fails (DL FAIL CORP) 291 76.13 -82.89 0.17 25.67 0.491

Panel D: Cost of CreditEffective Federal Funds Rate (D FF) 291 52.00 -102.00 0.47 10.41 1.002Federal Funds Target Rate (D TRGT) 291 25.00 -50.00 0.52 8.79 1.003Average Weekly Deviation From The Target (MISS) 291 45.00 -53.00 -0.10 6.73 0.042

Panel E: System Open Market AccountSOMA Repos (L FEDR) 292 4.11 1.95 3.27 0.29 0.338SOMA Reverse Repos (DL FEDRR) 215 26.08 -26.98 0.18 6.45 0.934SOMA All Treasuries (DL FEDTALL) 291 3.30 -2.23 0.13 0.32 0.996SOMA Treasury Bills (DL FEDBILL) 291 9.45 -7.20 0.15 0.89 0.993SOMA Fraction of Bills to Treasuries (DFEDALPHA) 291 0.02 -0.02 0.00 0.00 0.980

Panel F: Real Activity and Energy PricesWest Texas Intermediate Crude Oil (DL OIL) 291 8.70 -19.23 0.26 4.02 0.995Four Week Moving Average of Continuing Claims (DL CLAIMS) 291 7.75 -5.03 -0.08 1.70 0.992

Source: Authors’ calculations based on data from the Board of Governors H.15 release, Board of Governors press releases, theFederal Reserve Bank of New York FR 2004C report, the U.S. Department of Labor Unemployment Insurance Weekly ClaimsReport, and the U.S. Department of Energy Petroleum Status Report.

Note: The table reports univariate descriptive statistics of extreme values, central tendency (µ), dispersion (σ), and persistence(ρ) defined as the coefficient of a first-order autoregressive equation (in levels) for each column variable.

52

Table 1.3: Net Repo Borrowing and the Cost of Credit: July 4, 2001 to January 31, 2007Panel A: NETREPO

(1) (2) (1) (2) (1) (2)TRGT 123.84*** 125.54***

(15.31) (14.22)FF 123.88*** 125.66***

(15.51) (14.37)MISS -3.51** -4.32**

(1.69) (2.07)Constant 489.61*** 489.58*** 804.11***

(48.08) (48.38) (48.36)Fixed Effects N Y N Y N Y

R2 0.63 0.66 0.63 0.65 0.01 0.03F-test 12.35*** 11.55*** 3875.03***

Panel B: NETREPO OC

(1) (2) (1) (2) (1) (2)TRGT 137.35*** 138.90***

(14.97) (13.87)FF 137.42*** 139.05***

(15.18) (14.02)MISS -3.77** -4.59**

(1.77) (2.20)Constant 576.80*** 576.70*** 925.61***


R2 0.65 0.67 0.65 0.67 0.01 0.03F-test 18.18*** 17.65*** 4221.74***

Panel C: NETREPO TERM

(1) (2) (1) (2) (1) (2)TRGT -13.51*** -13.35***

(3.21) (2.95)FF -13.54*** -13.39***

(3.22) (2.95)MISS 0.24 0.27

(0.001) (0.34)Constant -87.19*** -87.12*** -121.50***


R2 0.25 0.36 0.26 0.36 0.001 0.11F-test 29.33*** 28.27*** 5795.53***

Obs. = 292

Note: The table reports OLS regressions of net repo borrowing by primary dealers on the cost of credit in the U.S. financialsystem. The dependent variable is alternatively net repo, net overnight and continuing repo, or net term repo. The costof credit is defined as either the effective federal funds rate, the federal funds target, or the average weekly deviation fromthe target. Newey-West (1987) standard errors allowing for 13 weeks of lags are in parenthesis. The F-test is of the jointsignificance of weekly fixed effects. *Statistically significant at the 10 percent level. **Statistically significant at the 5 percentlevel. ***Statistically significant at the 1 percent level.

53

Table 1.4: Net Financing and the Cost of Credit: July 4, 2001 to January 31, 2007Panel A: NETFINANCING

(1) (2) (1) (2) (1) (2)TRGT 13.01** 13.22***

(5.85) (5.13)FF 13.04** 13.30***

(5.86) (5.11)MISS -0.18 -0.04

(0.39) (0.51)Constant 190.66*** 190.55*** 233.70***


R2 0.15 0.31 0.15 0.31 0.0006 0.15F-test 24.45*** 24.32*** 3837.61***

Panel B: NET OC

(1) (2) (1) (2) (1) (2)TRGT 65.32*** 65.86***

(9.49) (8.02)FF 65.46*** 66.06***

(9.5) (7.99)MISS -1.28 -1.41

(0.82) (1.08)Constant 365.66*** 365.34*** 531.57***


R2 0.59 0.64 0.59 0.64 0.004 0.05F-test 11.61*** 11.73*** 4152.04***

Panel C: NET TERM

(1) (2) (1) (2) (1) (2)TRGT -52.31*** -52.64***

(6.14) (5.38)FF -52.41*** -52.76***

(6.19) (5.42)MISS 1.10* 1.37*

(0.63) (0.80)Constant -175.00*** -174.79*** -307.87.57***


R2 0.60 0.63 0.60 0.63 0.005 0.04F-test 11.61*** 9.20*** 3276.50***

Obs. = 292

Note: The table reports OLS regressions of net collateralized borrowing by primary dealers on the cost of credit in the U.S.financial system. The dependent variable is alternatively net financing, net overnight and continuing financing, or net termfinancing. The cost of credit is defined as either the effective federal funds rate, the federal funds target, or the average weeklydeviation from the target. Newey-West (1987) standard errors allowing for 13 weeks of lags are in parenthesis. The F-test is ofthe joint significance of weekly fixed effects. *Statistically significant at the 10 percent level. **Statistically significant at the5 percent level. ***Statistically significant at the 1 percent level.

54

Table 1.5: Gross Financing and Changes in the Cost of Credit: July 4, 2001 to January 31,2007

D FF D TRGT MISS(1) (2) (1) (2) (1) (2)

DL R -0.017 -0.080*** -0.021 -0.009 0.014 -0.072***DL OC R 0.072*** 0.072*** -0.000 0.002 -0.042 -0.020DL TERM R -0.132*** -0.200*** -0.006 0.009 0.109 -0.109**

DLALL OUT -0.021 -0.076*** -0.023 -0.010 0.010 -0.068***DLALL OUT OC 0.054*** 0.051*** -0.004 -0.004 -0.043 -0.021DLALL OUT TERM -0.120** -0.194*** -0.003 0.013 0.100 -0.110**Fixed Effects N Y N Y N Y

Obs. = 278

Note: The table reports impact multipliers for a 13 week autoregressive distributed lag model of the growth rate of grosscollateralized borrowing on the cost of credit in the U.S. financial system. The dependent variable is alternatively repo,overnight and continuing repo, term repo, securities out, overnight and continuing securities out, or term securities out. Thecost of credit is defined as either the effective federal funds rate, the federal funds target, or the average weekly deviation fromthe target. *Statistically significant at the 10 percent level. **Statistically significant at the 5 percent level. ***Statisticallysignificant at the 1 percent level.

55

Table 1.6: Daily Forecast Regressions of Federal Funds Target Rate Changes: July 4, 2001-January 31, 2007

D FF MISS D TRGT(1) (2) (1) (2) (1) (2)

1. Lag on D TRGT:

Intercept 0.001 0.001 0.001 0.001(0.001) (0.001) (0.70) (0.001)

β 0.003 0.003 0.063*** 0.052***(0.015) (0.015) (0.024) (0.019)

R2 0.00 0.00 0.03 0.02

2. Lag on D FF:

Intercept 0.001 0.001(0.002) (0.002)

β 0.363** 0.188***(0.163) (0.042)

R2 0.07 0.01

3. Lag on MISS:

Intercept 0.003 0.004**(0.002) (0.002)

β 0.392 0.002(0.360) (0.072)

R2 0.02 0.00

Obs. 2036 2035 2036 2035 2036 2035

Note: The table reports the results from OLS regressions of the column (dependent) on the row (independent) variable and aconstant. The second column of each regression excludes the September 17, 2001 change. Parenthesis contain standard errors.*Statistically significant at the 10 percent level. **Statistically significant at the 5 percent level. ***Statistically significant atthe 1 percent level.

Table 1.7: Innovations in The Reserve Market: July 4, 2001-January 31, 2007

εFF εRR εNBR

εFF 1.00 -0.07 -0.05εRR 1.00 0.94εNBR 1.00

Note: The table reports correlation coefficients between residuals obtained from a vector autoregression model with 12 lags.The VAR includes the following three variables (in levels): required reserves (RR), nonborrowed reserves (NBR), and the federalfunds rate (FF).

56

Table 1.8: First-Stage Results: εNBR Regressed on Instruments: July 4, 2001-January 31,2007

(1) (2)εRR 1.0767***

(0.0235)εDL CLAIMS -0.116*

(0.061)Constant -0.000 -0.000002

(0.0195) (0.063)R2 0.883 0.013Instrument F-test (statistic) 2107** 3.59

Obs. = 280

Note: The table reports first stage results from OLS regressions using residuals from a vector autoregression model with 12 lags.The VAR includes: required reserves (RR) or the four week moving average of initial jobless claims (DL CLAIMS), nonborrowedreserves (NBR), and the federal funds rate (FF). Standard errors in parenthesis. Significance of the F-test statistic is based onStock and Yugo (2005) 5% critical values for the test of weak instrument bias in a linear IV regression. *Statistically significantat the 10 percent level. **Statistically significant at the 5 percent level. ***Statistically significant at the 1 percent level.

Table 1.9: Estimated Slope of the Supply Function for Nonborrowed Reserves: July 4, 2001-January 31, 2007

OLS IV OLS IV(3) (4) (5) (6)

εNBR -0.003 0.0059(0.003) (0.0091)

εDL CLAIMS 0.0004 -0.000004(0.003) (0.003)

Constant -0.000 -0.000 -0.000 -0.000(0.003) (0.003) (0.003) (0.003)

R2 0.003 0.000 0.000 0.000Instrument F-test (statistic) 2107** 3.59Obs. 280 280 280 280

Note: The table reports second stage and raw results from OLS regressions using residuals from a vector autoregression modelwith 12 lags. The VAR includes: required reserves (RR) or the four week moving average of initial jobless claims (DL CLAIMS),nonborrowed reserves (NBR), and the federal funds rate (FF). Standard errors in parenthesis. F-test statistic significance fromthe first stage regression is based on Stock and Yugo (2005) 5% critical values for the test of weak instrument bias in a linearIV regression. *Statistically significant at the 10 percent level. **Statistically significant at the 5 percent level. ***Statisticallysignificant at the 1 percent level.

57

Table 1.10: Variance Decompositions of Gross Repo: July 4, 2001-January 31, 2007Horizon D FF D TRGT MISS L FEDR DL FEDRR DL FEDTALL DL FEDBILL DFEDALPHA

I. DL R13 Weeks 1.8 1.6 3.4 1.4 11.5 2.1 2.3 2.0

[0.6,6.2] [0.6,12.3] [0.6,12.3] [0.2,6.8] [3.2,29.1] [0.6,7.8] [0.8,7.4] [0.7,7.2]26 Weeks 2.1 2.2 4.1 2.9 12.8 2.7 2.9 2.6

[0.4,9.7] [0.4,10.9] [0.8,16.8] [0.3,12.9] [3.6,35.1] [0.6,12.1] [0.8,11.8] [0.7,7.2]39 Weeks 2.4 2.5 4.3 3.6 13.7 3.0 3.1 2.8

[0.3,12.5] [0.3,13.8] [0.7,18.5] [0.3,16.2] [3.5,37.7] [0.5,14.7] [0.6,14.1] [0.6,12.7]52 Weeks 2.7 2.8 4.4 3.9 14.6 3.1 3.2 2.9

[0.3,14.0] [0.3,15.6] [0.6,19.7] [0.3,17.4] [3.4,39.7] [0.4,16.4] [0.6,15.2] [0.5,13.7]

II. DL OC R13 Weeks 2.7 6.6 2.0 8.3 7.9 4.8 4.1 3.0

[0.8,8.8] [1.2,18.9] [0.6,7.1] [1.3,21.3] [3.6,16.4] [1.0,15.2] [1.1,13.6] [0.8,10.7]26 Weeks 3.5 9.9 2.4 9.7 7.4 7.9 5.4 3.5

[1.0,10.8] [1.4,28.7] [0.6,9.7] [1.2,27.9] [3.0,17.9] [1.1,24.2] [1.0,20.4] [0.7,15.1]39 Weeks 3.6 11.6 2.4 10.1 7.3 8.6 5.7 3.6

[0.9,11.7] [1.4,32.3] [0.5,10.8] [1.1,30.2] [2.4,19.1] [1.1,26.6] [0.9,22.6] [0.6,16.6]52 Weeks 3.4 12.5 2.4 10.2 7.1 9.0 5.8 3.6

[0.8,12.0] [1.5,34.1] [0.4,11.0] [1.0,31.1] [2.2,20.6] [1.0,27.9] [0.7,23.7] [0.5,17.3]

III. DL TERM R13 Weeks 2.4 2.0 3.9 1.0 12.3 2.0 2.1 2.0

[0.6,9.5] [0.5,8.8] [0.7,13.4] [0.2,4.7] [4.6,29.6] [0.6,7.4] [0.7,7.6] [0.7,6.8]26 Weeks 3.7 3.7 4.3 1.6 12.9 2.8 2.8 2.6

[0.4,16.6] [0.4,16.5] [0.8,17.8] [0.2,8.1] [4.2,35.6] [0.6,7.4] [0.7,13.0] [0.6,10.3]39 Weeks 4.4 4.7 4.2 1.9 13.9 3.2 3.1 2.7

[0.4,20.6] [0.4,21.2] [0.7,19.7] [0.2,10.2] [4.0,38.9] [0.6,15.8] [0.6,15.6] [0.5,11.9]52 Weeks 4.9 5.2 4.2 2.0 14.6 3.3 3.1 2.7

[0.3,23.1] [0.4,24.1] [0.6,20.8] [0.2,11.3] [3.8,40.5] [0.5,17.4] [0.5,17.0] [0.5,12.7]

Note: The table records the percentages of the variance of the forecasted variable accounted for by variation in the columnvariable at 13 week horizons. Point estimates are based on a vector autoregressions with 13 weekly lags of each variable includedin the VAR. The ordering of the variables in the variance decomposition is based on the benchmark identification: DL CLAIMS,the column variable, DL OIL, and the panels repo variable listed in the heading. Numbers in brackets represent 90 percentconfidence intervals. Numbers in bold are the largest percentage variance for each row.

58

Table 1.11: Variance Decompositions of Securities Out and Settlement Fails: July 4, 2001-January 31, 2007

Horizon D FF D TRGT MISS L FEDR DL FEDRR DL FEDTALL DL FEDBILL DFEDALPHA

I. DLALL OUT13 Weeks 1.7 1.7 3.5 1.4 10.8 2.0 2.4 2.1

[0.6,6.4] [0.5,6.5] [0.6,12.8] [0.2,7.5] [3.2,27.7] [0.6,6.9] [0.8,7.4] [0.7,6.9]26 Weeks 2.2 2.3 4.2 2.8 12.3 2.5 2.9 2.8

[0.4,10.1] [0.5,11.1] [0.8,17.4] [0.3,13.9] [3.8,34.5] [0.5,11.5] [0.7,12.5] [0.7,11.5]39 Weeks 2.6 2.8 4.6 3.5 13.4 2.7 3.2 3.1

[0.3,12.8] [0.4,14.2] [0.8,19.3] [0.3,17.9] [3.7,38.6] [0.5,14.0] [0.6,15.5] [0.6,13.7]52 Weeks 2.8 3.1 4.6 3.9 13.8 2.9 3.2 3.2

[0.3,14.4] [0.3,16.0] [0.7,20.7] [0.3,19.5] [3.6,40.3] [0.4,15.9] [0.6,17.3] [0.5,14.9]

II. DLALL OUT OC13 Weeks 5.7 7.3 2.0 6.9 7.1 3.6 3.2 2.5

[1.1,16.1] [1.3,19.8] [0.6,8.1] [0.9,19.4] [3.1,15.5] [0.8,12.4] [0.9,10.7] [0.8,8.5]26 Weeks 8.2 11.3 6.2 4.2 7.2 5.6 4.0 2.8

[1.0,23.3] [1.4,29.5] [1.9,17.6] [1.4,12.4] [2.7,17.4] [0.8,19.8] [0.8,16.1] [0.6,11.8]39 Weeks 9.0 12.5 6.0 4.1 7.0 6.1 4.2 2.8

[1.0,26.4] [1.5,33.4] [1.6,18.8] [1.3,13.6] [2.3,19.0] [0.7,21.4] [0.7,17.8] [0.5,13.0]52 Weeks 9.5 13.2 5.9 3.9 7.1 6.1 4.2 2.8

[1.0,27.9] [1.5,35.6] [1.4,19.5] [1.1,14.2] [2.2,20.1] [0.6,22.3] [0.6,18.4] [0.4,13.4]

III. DLALL OUT TERM13 Weeks 2.2 1.7 4.3 0.9 13.3 1.9 2.1 2.1

[0.6,8.9] [0.5,7.5] [0.7,14.2] [0.2,4.6] [5.1,30.7] [0.6,6.9] [0.7,7.4] [0.7,6.8]26 Weeks 3.2 3.0 4.4 1.6 14.3 2.8 2.9 2.5

[0.4,15.6] [0.4,15.5] [0.9,18.0] [0.2,8.4] [4.6,37.1] [0.7,11.8] [0.8,12.7] [0.6,10.1]39 Weeks 3.9 3.8 4.4 2.0 15.1 3.1 3.1 2.5

[0.4,19.1] [0.4,19.7] [0.8,19.8] [0.2,10.7] [4.1,41.3] [0.6,14.7] [0.6,14.9] [0.5,11.8]52 Weeks 4.2 4.3 4.3 2.2 15.9 3.3 3.2 2.5

[0.3,21.1] [0.3,22.1] [0.7,20.5] [0.2,11.5] [3.8,42.4] [0.5,16.3] [0.5,16.2] [0.5,12.8]

IV. DLALL FAIL13 Weeks 6.4 2.4 3.4 1.7 5.4 2.0 3.1 3.5

[1.5,16.6] [0.7,9.8] [0.8,11.0] [0.3,7.0] [2.1,14.9] [0.7,7.0] [1.1,8.5] [1.2,9.1]26 Weeks 5.3 2.5 6.2 4.1 7.2 2.2 3.0 3.6

[0.9,18.4] [0.4,12.1] [1.0,19.5] [0.4,15.9] [2.3,22.8] [0.5,9.7] [0.8,11.5] [0.9,12.7]39 Weeks 4.9 2.7 6.8 5.2 7.6 2.3 3.1 3.9

[0.6,19.9] [0.3,13.5] [0.9,22.1] [0.4,20.2] [2.2,24.9] [0.4,11.3] [0.6,13.8] [0.7,14.9]52 Weeks 4.7 2.7 7.1 5.7 8.0 2.3 3.1 4.0

[0.5,20.4] [0.3,14.4] [0.8,23.0] [0.4,22.1] [2.1,26.3] [0.4,12.4] [0.5,14.8] [0.6,16.2]

Note: The table records the percentages of the variance of the forecasted variable accounted for by variation in the columnvariable at 13 week horizons. Point estimates are based on a vector autoregressions with 13 weekly lags of each variable includedin the VAR. The ordering of the variables in the variance decomposition is based on the benchmark identification: DL CLAIMS,the column variable, DL OIL, and the panels financing measure listed in the heading. Numbers in brackets represent 90 percentconfidence intervals. Numbers in bold are the largest percentage variance for each row.

59

Figure 1.1: System Open Market Account Holdings: December 18, 2002-January 31, 2007

SOM

A D

omes

tic P

ortf

olio

(bill

ions

) Holdings Held Outright by Security Class

WEEK2003 2004 2005 2006

200

300

400

500

600

700

800

ALL_US

ALL_BILLS

ALL_TNOT EBOND

Holdings Held Outright by Maturity

WEEK2003 2004 2005 2006

25

50

75

100

125

150

175

200

225

ALL_US15D

ALL_US16TO90D

ALL_US91TO365D

ALL_US1TO5Y

ALL_US5T O10Y

ALL_USOVER10Y

Temporary Holdings by Maturity

TEMP_OMO FEDR15D FEDRR15D

WEEK

(in b

illio

ns)

2003 2004 2005 200610

20

30

40

50

60

70

80

Source: Authors’ calculations based on data from the Board of Governors H.4.1 release.

Note: The figure plots outright and temporary holding of the System Open Market Account (in billions) by: Security class (topleft), security class by maturity (top right), and temporary holdings (bottom).

60

Figure 1.2: Temporary Open Market Operation - Tri-party Repo

CustodianBank

PrimaryDealer

(Borrower)

Monetary Authority(Lender)

Bond

Reserves Reserves

T = 0

CustodianBank

PrimaryDealer

(Borrower)

Monetary Authority(Lender)

Bond

Reserves Reserves

T = 1

Note: A Fed repo is an open market operation which temporarily adds reserves to the banking system. It is conducted viaauction at the initiative of the open market trading desk of the FRBNY. At T = 0, through FedTrade each dealer is requestedto present the rates they are willing to pay to borrow money against Treasury, agency, and mortgage backed collateral. Winningbids are selected on a competitive basis. The dealer delivers collateral to the Fed’s custodial account at the dealers tri-partyagent, also known as the clearing bank. The Fed makes payment by crediting the reserve account of the clearing bank. Whenthe repo matures at T = 1, the dealer returns the loan plus interest (repurchase price) and the tri-party agent returns thecollateral.

61

Figure 1.3: Temporary Open Market Operation - Bilateral Reverse Repo

MMF(Lender)

Monetary Authority

(Borrower)

Reserves

Bond

T = 0

MMF(Lender)

Monetary Authority

(Borrower)

Reserves

Bond

T = 1

Note: A Fed reverse repo is an open market operation which temporarily drains reserves from the banking system. It is conductedvia auction at the initiative of the open market trading desk of the FRBNY. At T = 0, through FedTrade each participant isrequested to offer the rates they are willing to lend money to the Fed versus Treasury collateral. Winning offers are selected ona competitive basis. Settlement is delivery versus payment so the delivery of collateral and reserves is simultaneous. When thedeal matures at T = 1, the lender and the Fed return the collateral and the loan plus interest (repurchase price) respectively.

62

Figure 1.4: Gross Repo Activity: July 4, 2001-January 31, 2007

Gross Repo vs. Gross Financing

ALL_R ALL_OUT

YEAR2001 2002 2003 2004 2005 2006

1500

2000

2500

3000

3500

4000

4500

Gross Securities Out vs. Gross Securities In

ALL_OUT ALL_IN ALL_RR

YEAR2001 2002 2003 2004 2005 2006

1000

1500

2000

2500

3000

3500

4000

4500

Overnight Gross Repo vs. Overnight Gross Financing

OPEN_REPO ALL_OUT_OC

YEAR2001 2002 2003 2004 2005 2006

750

1000

1250

1500

1750

2000

2250

2500

2750

Term Gross Repo vs. Term Gross Financing

TERM_REPO ALL_OUT_TERM

YEAR2001 2002 2003 2004 2005 2006

600

800

1000

1200

1400

1600

1800

2000

Source: Authors’ calculations based on data from the Federal Reserve’s FR 2004C Report.

Note: The figure plots weekly gross repo and gross financing (in billions of dollars) by maturity. The bottom leftpanel plots gross financing for securities out, securities in, and gross reverse repo financing reported by the primary dealers.The shading reflects NBER business cycle dates.

63

Figure 1.5: Primary Dealers Net Financing By Maturity: July 4, 2001-January 31, 2007

Net Repo vs. Net Financing

YEAR2001 2002 2003 2004 2005 2006

0

200

400

600

800

1000

1200

1400

NETREPO

NETFINANCIN G

Net Repo Financing

YEAR2001 2002 2003 2004 2005 2006

-250

0

250

500

750

1000

1250

1500

NETREPO

NETREPO_OPEN

NETREPO_TERM

Overnight Net Repo vs. Overnight Net Financing

YEAR2001 2002 2003 2004 2005 2006

250

500

750

1000

1250

1500

NETREPO_OPEN

N ET_OPEN

Term Net Repo vs. Term Net Financing

YEAR2001 2002 2003 2004 2005 2006

-600

-500

-400

-300

-200

-100

0

100

NETREPO_TERM

NET_TER M


Note: The figure plots weekly net financing and net repo financing (in billions of dollars) by maturity. Net financing by dealersis calculated as securities delivered (out) minus securities received (in). Net repo financing by dealers is calculated as repominus reverse repo. The shading reflects NBER business cycle dates.

64

Figure 1.6: Primary Dealers Net Financing and Fails By Security Class: July 4, 2001-January31, 2007

Net Financing By Security Class

YEAR2001 2002 2003 2004 2005 2006

-300

-200

-100

0

100

200

300

400

NET_U S

NET_AGENC Y

NET_MBS NET_C ORP

Financing Fails By Security Class

FAIL_US FAIL_MBS FAIL_AGENCY

YEAR2001 2002 2003 2004 2005 2006

0

250

500

750

1000

1250

1500

1750

Overnight Net Financing By Security Class

YEAR2001 2002 2003 2004 2005 2006

-200

-100

0

100

200

300

400

500

NET_OC_U S

NET_OC_AGENC Y

NET_OC_MBS

NET_OC_CORP

Term Net Financing By Security Class

YEAR2001 2002 2003 2004 2005 2006

-250

-200

-150

-100

-50

0

50

NET_TER M_U S

NET_TER M_AGEN CY

NET_TERM_MBS

NET_TERM_CORP


Note: The figure plots weekly cumulative fails and net financing by security class (in billions of dollars). Net financing bydealers is securities delivered (out) minus securities received (in). The shading marks NBER business cycle dates.

65

Figure 1.7: Monetary Policy and Securitized Banking

Obligors

SPV

PrimaryDealerHedge fund

MMF

Monetary Authority

A(See Fig. 1)

B(See Fig. 2)

Note: The figure maps short term funding flows from the monetary authority to activity in the shadow banking system. Themonetary authority conducts open market operations with primary dealers and money market mutual funds which directlyimpacts the extension of credit in the shadow banking system.

66

Figure 1.8: Measures of Real Activity: July 4, 2001-January 31, 2007

Source: Authors’ calculations based on Federal Reserve Economic Data (FRED).

Note: The figure plots the four week moving average of initial jobless claims (top and middle) and the WTI crude oil spot price(bottom) with alternate monthly and quarterly economic indicators.

67

Figure 1.9: The Federal Funds Rate and Federal Funds Target in 2001 and 2004

Inte

rest

Rat

e (p

erce

nt)

Easing

FF TRGT

DAY

J A S O N D J F M A M J2001 2002

1.0

1.5

2.0

2.5

3.0

3.5

4.0Tightening

FF TRGT

DAY

J J A S O N D J F M A M2004

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Source: Authors’ calculations based on data from the Board of Govenors H.15 release.

Note: The figure plots the daily effective federal funds rate and target rate for one year during a monetary easing (July 4,2001-July 4, 2002) and tightening (June 1, 2004-June 1, 2004) cycle. The shading reflects NBER business cycle dates.

68

Figure 1.10: Response of The Federal Funds Rate to Energy and Employment Shocks

Shock to DL CLAIMS Shock to DL OIL

DL

CL

AIM

S

0 5 10 15 20 25 30 35 40 45 500.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5 DL_CLAIMS ‐> DLCLAIMS

Week

Percent

0 5 10 15 20 25 30 35 40 45 50-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

Percent

DL_OIL ‐> DL_CLAIMS

Week

DF

F

0 5 10 15 20 25 30 35 40 45 50-10.0

-7.5

-5.0

-2.5

0.0

2.5

Percent

DL_CLAIMS ‐> D_FF

Week0 5 10 15 20 25 30 35 40 45 50

-2

0

2

4

6

8

10

Percent

DL_OIL ‐> D_FF

Week

DL

OIL

0 5 10 15 20 25 30 35 40 45 50-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Percent

DL_CLAIMS ‐> DL_OIL

Week0 5 10 15 20 25 30 35 40 45 50

0

1

2

3

4

5

6

Percent

DL_OIL ‐> DL_OIL

Week

Note: The figure plots impulse responses of the federal funds rate to energy and employment shocks from a recursive VARmodel based on the benchmark ordering. The black line is the median of the simulated responses, the blue line represents 90%probability bands, and the green line represents 84% probability bands.

69

Figure 1.11: The Response of Open Market Operations to Reserve Imbalance Shocks

Shock to MISSTemporary Permanent

0 5 10 15 20 25 30 35 40 45 50-0.2

0.0

0.2

0.4

0.6

0.8

1.0 MISS ‐> L_FEDR

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

0.25 MISS ‐> DL_FEDTALL

Percent

Week

L FEDR DL FEDTALL

0 5 10 15 20 25 30 35 40 45 50-7

-6

-5

-4

-3

-2

-1

0 MISS ‐> DL_FEDRR

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6 MISS ‐> DL_FEDBILL

Percent

Week

DL FEDRR DL FEDBILL

0 5 10 15 20 25 30 35 40 45 50-0.00050

-0.00025

0.00000

0.00025

0.00050

0.00075

0.00100

0.00125

0.00150 MISS ‐> DFEDALPHA

Percent

Week

DFEDALPHA

Note: The figure plots impulse responses of system open market account holdings to average weekly target deviation shocksfrom a bivariate recursive VAR model with MISS ordered first. The black line is the median of the simulated responses, theblue line represents 90% probability bands, and the green line represents 68% probability bands.

70

Figure 1.12: Responses to a Monetary Policy Shock

Shock to

0 5 10 15 20 25 30 35 40 45 500.0

2.5

5.0

7.5

10.0

12.5

15.0

17.5

20.0

22.5

Percent

D_FF ‐> D_FF

Week0 5 10 15 20 25 30 35 40 45 50

0.0

2.5

5.0

7.5

10.0

12.5

15.0

17.5

20.0

22.5

Percent

D_TRGT ‐> D_TRGT

Week0 5 10 15 20 25 30 35 40 45 50

0

2

4

6

8

10

12

Percent

MISS ‐> MISS

Week

D FF D TRGT MISS

0 5 10 15 20 25 30 35 40 45 500.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Percent

L_FEDR ‐> L_FEDR

Week0 5 10 15 20 25 30 35 40 45 50

0

1

2

3

4

5

6

7

Percent

DL_FEDRR ‐> DL_FEDRR

Week

L FEDR DL FEDRR

0 5 10 15 20 25 30 35 40 45 500.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Percent

DL_FEDTALL ‐> DL_FEDTALL

Week0 5 10 15 20 25 30 35 40 45 50

0.00

0.25

0.50

0.75

1.00

1.25

Percent

Week

DL_FEDBILL ‐> DL_FEDBILL

0 5 10 15 20 25 30 35 40 45 500.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0030

0.0035 DFEDALPHA ‐> DFEDALPHA

Percent

Week

DL FEDTALL DL FEDBILL DFEDALPHA

Note: The figure plots impulse responses of monetary policy instruments to monetary policy instrument shocks from a recursiveVAR model based on the modified benchmark ordering. The black line is the median of the simulated responses, the blue linerepresents 90% probability bands, and the green line represents 84% probability bands.

71

Figure 1.13: Response of Real Activity to a Shock to the Federal Funds Rate

Shock to DL DFF

DL

CL

AIM

S

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

Percent

D_FF ‐> DL_CLAIMS

Week

DL

OIL

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Percent

D_FF ‐> DL_OIL

Week

Note: The figure plots impulse responses of the four week average of initial jobless claims and the WTI spot price of crude oilto federal funds rate shocks from a recursive VAR model based on the benchmark ordering. The black line is the median of thesimulated responses, the blue line represents 90% probability bands, and the green line represents 84% probability bands.

72

Figure 1.14: Response of Repo to Cost of Credit Shocks

Shock to D FF Shock to D TRGT Shock to MISS

DL

R

0 5 10 15 20 25 30 35 40 45 50-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

Percent

D_FF ‐> DL_R

Week0 5 10 15 20 25 30 35 40 45 50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

Percent

D_TRGT ‐> DL_R

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Week

MISS ‐> DL_R

Percent

DL

OC

R

0 5 10 15 20 25 30 35 40 45 50-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

Percent

D_FF ‐> DL_OPEN_R

Week0 5 10 15 20 25 30 35 40 45 50

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

Percent

D_TRGT ‐> DL_OPEN_R

Week0 5 10 15 20 25 30 35 40 45 50

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4

Week

MISS ‐> DL_OPEN_R

Percent

DL

TE

RM

R

0 5 10 15 20 25 30 35 40 45 50-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

Percent

D_FF ‐> DL_TERM_R

Week0 5 10 15 20 25 30 35 40 45 50

-5

-4

-3

-2

-1

0

1

Percent

D_TRGT ‐> DL_TERM_R

Week0 5 10 15 20 25 30 35 40 45 50

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Week

MISS ‐> DL_TERM_R

Percent

Note: The figure plots impulse responses of repo financing to monetary policy shocks from a recursive VAR model based on themodified benchmark ordering. The black line is the median of the simulated responses, the blue line represents 90% probabilitybands, and the green line represents 84% probability bands.

73

Figure 1.15: Response of Securities Out to Cost of Credit Shocks


DL

AL

LO

UT

R

0 5 10 15 20 25 30 35 40 45 50-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

Percent

D_FF ‐> DLALL_OUT

Week0 5 10 15 20 25 30 35 40 45 50

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

Percent

D_TRGT ‐> DLALL_OUT

Week0 5 10 15 20 25 30 35 40 45 50

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Week

MISS ‐> DLALL_OUT

Percent

DL

AL

LO

UT

OC

R

Percent

D_FF ‐> DLALL_OUT_OC

Week0 5 10 15 20 25 30 35 40 45 50

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

Percent

D_TRGT ‐> DLALL_OUT_OC

Week0 5 10 15 20 25 30 35 40 45 50

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

Week

MISS ‐> DLALL_OUT_OC

Percent

0 5 10 15 20 25 30 35 40 45 50-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4

DL

AL

LO

UT

TE

RM

R

Percent

D_FF ‐> DLALL_OUT_TERM

Week0 5 10 15 20 25 30 35 40 45 50

-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

Percent

D_TRGT ‐> DLALL_OUT_TERM

Week0 5 10 15 20 25 30 35 40 45 50

-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Week

MISS ‐> DLALL_OUT_TERM

Percent

0 5 10 15 20 25 30 35 40 45 50-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Note: The figure plots impulse responses of securities out to monetary policy shocks from a recursive VAR model based on themodified benchmark ordering. The black line is the median of the simulated responses, the blue line represents 90% probabilitybands, and the green line represents 84% probability bands.

74

Figure 1.16: Response of Financing Fails to Cost of Credit Shocks


0 5 10 15 20 25 30 35 40 45 50-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

Percent

D_FF ‐> DLALL_FAIL

Week0 5 10 15 20 25 30 35 40 45 50

-12.5

-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

Percent

D_TRGT ‐> DLALL_FAIL

Week0 5 10 15 20 25 30 35 40 45 50

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

12.5

15.0

Week

MISS ‐> DLALL_FAIL

Percent

Shock to L FEDR Shock to DL FEDRR

0 5 10 15 20 25 30 35 40 45 50-25

-20

-15

-10

-5

0

5

Percent

L_FEDR ‐> DLALL_FAIL

0 5 10 15 20 25 30 35 40 45 50-15

-10

-5

0

5

10

15

20

25 DL_FEDRR ‐> DLALL_FAIL

Percent

Week

Shock to DL FEDTALL Shock to DL FEDBILL Shock to DFEDALPHA

0 5 10 15 20 25 30 35 40 45 50-15

-10

-5

0

5

10

15

Week

Percent

DL_FEDTALL ‐> DLALL_FAIL DL_FEDBILL ‐> DLALL_FAIL

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-5

0

5

10

15

20

25

0 5 10 15 20 25 30 35 40 45 50-5

0

5

10

15

20

25

30

Percent

DFEDALPHA ‐> DLALL_FAIL

Week

Note: The figure plots impulse responses of financing fails to monetary policy shocks from a recursive VAR model based on themodified benchmark ordering. The black line is the median of the simulated responses, the blue line represents 90% probabilitybands, and the green line represents 84% probability bands.

75

Figure 1.17: Response of Securities Out By Collateral Class to Cost of Credit Shocks


US

0 5 10 15 20 25 30 35 40 45 50-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

D_FF ‐> DLALL_OUT_US

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

D_TRGT ‐> DLALL_OUT_US

Week0 5 10 15 20 25 30 35 40 45 50

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Week

MISS ‐> DLALL_OUT_US

Percent

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

D_FF ‐> DLALL_OUT_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

D_TRGT ‐> DLALL_OUT_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Week

MISS ‐> DLALL_OUT_AGENCY

Percent

MB

S

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4

0.6

Percent

D_FF ‐> DLALL_OUT_MBS

Week0 5 10 15 20 25 30 35 40 45 50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

D_TRGT ‐> DLALL_OUT_MBS

Week0 5 10 15 20 25 30 35 40 45 50

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Week

MISS ‐> DLALL_OUT_MBS

Percent

CO

RP

0 5 10 15 20 25 30 35 40 45 50-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

Percent

D_FF ‐> DLALL_OUT_CORP

Week0 5 10 15 20 25 30 35 40 45 50

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4

0.6

Percent

D_TRGT ‐> DLALL_OUT_CORP

Week0 5 10 15 20 25 30 35 40 45 50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

Week

MISS ‐> DLALL_OUT_CORP

Percent

Note: The figure plots impulse responses of collateralized financing by collateral class to monetary policy shocks from a recursiveVAR model based on the modified benchmark ordering. The black line is the median of the simulated responses, the blue linerepresents 90% probability bands, and the green line represents 84% probability bands.

76

Figure 1.18: Response of Overnight and Continuing Securities Out By Collateral Class toCost of Credit Shocks

Shock to D FF Shock to D TRGT Shock to MISSU

S

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Percent

D_FF ‐> DLOUT_OC_US

Week0 5 10 15 20 25 30 35 40 45 50

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

Percent

D_TRGT ‐> DLOUT_OC_US

Week0 5 10 15 20 25 30 35 40 45 50

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

Week

MISS ‐> DLOUT_OC_US

Percent

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4

0.6

0.8

1.0

Percent

D_FF ‐> DLOUT_OC_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4

0.6

0.8

Percent

D_TRGT ‐> DLOUT_OC_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

Week

MISS ‐> DLOUT_OC_AGENCY

Percent

MB

S

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Percent

D_FF ‐> DLOUT_OC_MBS

Week0 5 10 15 20 25 30 35 40 45 50

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Percent

Week

D_TRGT ‐> DLOUT_OC_MBS

0 5 10 15 20 25 30 35 40 45 50-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Week

MISS ‐> DLOUT_OC_MBS

Percent

CO

RP

0 5 10 15 20 25 30 35 40 45 50-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

Percent

D_FF ‐> DLOUT_OC_CORP

Week0 5 10 15 20 25 30 35 40 45 50

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4

Percent

Week

D_TRGT ‐> DLOUT_OC_CORP

0 5 10 15 20 25 30 35 40 45 50-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

Week

MISS ‐> DLOUT_OC_CORP

Percent

Note: The figure plots impulse responses of overnight and continuing collateralized financing by collateral class to monetarypolicy shocks from a recursive VAR model based on the modified benchmark ordering. The black line is the median of thesimulated responses, the blue line represents 90% probability bands, and the green line represents 84% probability bands.

77

Figure 1.19: Response of Term Securities Out By Collateral Class to Cost of Credit Shocks


US

0 5 10 15 20 25 30 35 40 45 50-4

-3

-2

-1

0

1

Percent

D_FF ‐> DLOUT_TERM_US

Week0 5 10 15 20 25 30 35 40 45 50

-4

-3

-2

-1

0

1

Percent

Week

D_TRGT ‐> DLOUT_TERM_US

0 5 10 15 20 25 30 35 40 45 50-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Week

MISS ‐> DLOUT_TERM_US

Percent

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Percent

D_FF ‐> DLOUT_TERM_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

Percent

D_TRGT ‐> DLOUT_TERM_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-4

-3

-2

-1

0

1

2

3

Week

MISS ‐> DLOUT_TERM_AGENCY

Percent

MB

S

0 5 10 15 20 25 30 35 40 45 50-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

Percent

D_FF ‐> DLOUT_TERM_MBS

Week0 5 10 15 20 25 30 35 40 45 50

-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Percent

D_TRGT ‐> DLOUT_TERM_MBS

Week0 5 10 15 20 25 30 35 40 45 50

-4

-3

-2

-1

0

1

2

Week

MISS ‐> DLOUT_TERM_MBS

Percent

CO

RP

0 5 10 15 20 25 30 35 40 45 50-4

-3

-2

-1

0

1

2

3

Percent

D_FF ‐> DLOUT_TERM_CORP

Week0 5 10 15 20 25 30 35 40 45 50

-4

-3

-2

-1

0

1

2

3

Percent

D_TRGT ‐> DLOUT_TERM_CORP

Week0 5 10 15 20 25 30 35 40 45 50

-3

-2

-1

0

1

2

3

Week

MISS ‐> DLOUT_TERM_CORP

Percent

Note: The figure plots impulse responses of term collateralized financing by collateral class to monetary policy shocks from arecursive VAR model based on the modified benchmark ordering. The black line is the median of the simulated responses, theblue line represents 90% probability bands, and the green line represents 84% probability bands.

78

Figure 1.20: Response of Financing Fails By Collateral Class to Cost of Credit Shocks


US

0 5 10 15 20 25 30 35 40 45 50-20

-15

-10

-5

0

5

10

15

20

Percent

D_FF ‐> DLFAIL_US

Week0 5 10 15 20 25 30 35 40 45 50

-20

-15

-10

-5

0

5

10

15

20

Percent

D_TRGT ‐> DLFAIL_US

Week0 5 10 15 20 25 30 35 40 45 50

-10

-5

0

5

10

15

20

25

30

Week

MISS ‐> DLFAIL_US

Percent

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

12.5

15.0

Percent

D_FF ‐> DLFAIL_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

12.5

15.0

Percent

D_TRGT ‐> DLFAIL_AGENCY

Week0 5 10 15 20 25 30 35 40 45 50

-15

-10

-5

0

5

10

15

Week

MISS ‐> DLFAIL_AGENCY

Percent

MB

S

0 5 10 15 20 25 30 35 40 45 50-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

Percent

D_FF ‐> DLFAIL_MBS

Week0 5 10 15 20 25 30 35 40 45 50

-12.5

-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

Percent

D_TRGT ‐> DLFAIL_MBS

Week0 5 10 15 20 25 30 35 40 45 50

-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

10.0

12.5

Week

MISS ‐> DLFAIL_MBS

Percent

CO

RP

0 5 10 15 20 25 30 35 40 45 50-4

-3

-2

-1

0

1

2

3

4

Percent

D_FF ‐> DLFAIL_CORP

Week0 5 10 15 20 25 30 35 40 45 50

-6

-5

-4

-3

-2

-1

0

1

2

3

Percent

D_TRGT ‐> DLFAIL_US

Week0 5 10 15 20 25 30 35 40 45 50

-4

-2

0

2

4

6

8

Week

MISS ‐> DLFAIL_CORP

Percent

Note: The figure plots impulse responses of financing fails by collateral class to monetary policy shocks from a recursive VARmodel based on the modified benchmark ordering. The black line is the median of the simulated responses, the blue linerepresents 90% probability bands, and the green line represents 84% probability bands.

79

Figure 1.21: Response of Repo to Temporary Open Market Operation Shocks


DL

R

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Percent

L_FEDR ‐> DL_R

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0 DL_FEDRR ‐> DL_R

Percent

Week

DL

OC

R

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

Percent

L_FEDR ‐> DL_OPEN_R

0 5 10 15 20 25 30 35 40 45 50-0.5

0.0

0.5

1.0

1.5

2.0 DL_FEDRR ‐> DL_OPEN_R

Percent

Week

DL

TE

RM

R

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

Percent

L_FEDR ‐> DL_TERM_R

0 5 10 15 20 25 30 35 40 45 50-6

-5

-4

-3

-2

-1

0

1 DL_FEDRR ‐> DL_TERM_R

Percent

Week

Note: The figure plots impulse responses of repo financing to temporary liquidity injection and withdrawal shocks from arecursive VAR model based on the modified benchmark ordering. The black line is the median of the simulated responses, theblue line represents 90% probability bands, and the green line represents 84% probability bands.

80

Figure 1.22: Response of Securities Out By Collateral Class to Temporary Open MarketOperation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Percent

L_FEDR ‐> DL_OUT_US

0 5 10 15 20 25 30 35 40 45 50-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5 DL_FEDRR ‐> DLALL_OUT_US

Percent

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

Percent

L_FEDR ‐> DL_OUT_AGENCY

0 5 10 15 20 25 30 35 40 45 50-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00 DL_FEDRR ‐> DLALL_OUT_AGENCY

Percent

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

Percent

L_FEDR ‐> DLALL_OUT_MBS

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0 DL_FEDRR ‐> DLALL_OUT_MBS

Percent

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

Percent

L_FEDR ‐> DLALL_OUT_CORP

0 5 10 15 20 25 30 35 40 45 50-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0 DL_FEDRR ‐> DLALL_OUT_CORP

Percent

Week

Note: The figure plots impulse responses of collateralized financing by collateral class to temporary liquidity injection andwithdrawal shocks from a recursive VAR model based on the modified benchmark ordering. The black line is the median of thesimulated responses, the blue line represents 90% probability bands, and the green line represents 84% probability bands.

81

Figure 1.23: Response of Overnight and Continuing Securities Out By Collateral Class toTemporary Open Market Operation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0

Percent

L_FEDR ‐> DLOUT_OC_US

0 5 10 15 20 25 30 35 40 45 50-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0 DL_FEDRR ‐> DLOUT_OC_US

Percent

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Percent

L_FEDR ‐> DLOUT_OC_AGENCY

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0 DL_FEDRR ‐> DLOUT_OC_AGENCY

Percent

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Percent

L_FEDR ‐> DLOUT_OC_MBS

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5 DL_FEDRR ‐> DLOUT_OC_MBS

Percent

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-0.5

0.0

0.5

1.0

1.5

2.0

2.5

Percent

L_FEDR ‐> DLOUT_CORP

0 5 10 15 20 25 30 35 40 45 50-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75 DL_FEDRR ‐> DLOUT_OC_CORP

Percent

Week

Note: The figure plots impulse responses of overnight and continuing collateralized financing by collateral class to temporaryliquidity injection and withdrawal shocks from a recursive VAR model based on the modified benchmark ordering. The blackline is the median of the simulated responses, the blue line represents 90% probability bands, and the green line represents 84%probability bands.

82

Figure 1.24: Responses of Term Securities Out By Collateral Class to Temporary OpenMarket Operation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-3

-2

-1

0

1

2

3

4

Percent

L_FEDR ‐> DLOUT_TERM_US

0 5 10 15 20 25 30 35 40 45 50-7

-6

-5

-4

-3

-2

-1

0

1 DL_FEDRR ‐> DLOUT_TERM_US

Percent

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

5

6

Percent

L_FEDR ‐> DLOUT_TERM_AGENCY

0 5 10 15 20 25 30 35 40 45 50-4

-3

-2

-1

0

1

2

3 DL_FEDRR ‐> DLOUT_TERM_AGENCY

Percent

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

5

Percent

L_FEDR ‐> DLOUT_TERM_MBS

0 5 10 15 20 25 30 35 40 45 50-3

-2

-1

0

1

2

3

4 DL_FEDRR ‐> DLOUT_TERM_MBS

Percent

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-3

-2

-1

0

1

2

3

4

5

Percent

L_FEDR ‐> DLOUT_TERM_CORP

0 5 10 15 20 25 30 35 40 45 50-7.5

-5.0

-2.5

0.0

2.5 DL_FEDRR ‐> DLOUT_TERM_CORP

Percent

Week

Note: The figure plots impulse responses of term collateralized financing by collateral class to temporary liquidity injection andwithdrawal shocks from a recursive VAR model based on the modified benchmark ordering. The black line is the median of thesimulated responses, the blue line represents 90% probability bands, and the green line represents 84% probability bands.

83

Figure 1.25: Response of Financing Fails By Collateral Class to Temporary Open MarketOperation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-30

-20

-10

0

10

20

Percent

L_FEDR ‐> DLFAIL_US

0 5 10 15 20 25 30 35 40 45 50-30

-20

-10

0

10

20

30

40 DL_FEDRR ‐> DLFAIL_US

Percent

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-25

-20

-15

-10

-5

0

5

10

Percent

L_FEDR ‐> DLFAIL_AGENCY

0 5 10 15 20 25 30 35 40 45 50-10

0

10

20

30

40

50 DL_FEDRR ‐> DLFAIL_AGENCY

Percent

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-25

-20

-15

-10

-5

0

5

10

Percent

L_FEDR ‐> DLFAIL_MBS

0 5 10 15 20 25 30 35 40 45 50-10

-5

0

5

10

15

20

25 DL_FEDRR ‐> DLFAIL_MBS

Percent

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-5

-4

-3

-2

-1

0

1

2

3

4

Percent

L_FEDR ‐> DLFAIL_CORP

0 5 10 15 20 25 30 35 40 45 50-7.5

-5.0

-2.5

0.0

2.5

5.0 DL_FEDRR ‐> DLFAIL_CORP

Percent

Week

Note: The figure plots impulse responses of financing fails by collateral class to temporary liquidity injection and withdrawalshocks from a recursive VAR model based on the modified benchmark ordering. The black line is the median of the simulatedresponses, the blue line represents 90% probability bands, and the green line represents 84% probability bands.

84

Figure 1.26: Response of Repo to Permanent Open Market Operation Shocks


DL

R

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Week

Percent

DL_FEDTALL ‐> DL_R

0 5 10 15 20 25 30 35 40 45 50-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

0.2

0.4 DL_FEDBILL ‐> DL_R

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.6

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

-0.0

Percent

DFEDALPHA ‐> DL_R

Week

DL

OC

R

0 5 10 15 20 25 30 35 40 45 50-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

Week

Percent

DL_FEDTALL ‐> DL_OPEN_R

0 5 10 15 20 25 30 35 40 45 50-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00 DL_FEDBILL ‐> DL_OPEN_R

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

Percent

DFEDALPHA ‐> DL_OPEN_R

Week

DL

TE

RM

R

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

Week

Percent

DL_FEDTALL ‐> DL_TERM_R

0 5 10 15 20 25 30 35 40 45 50-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5 DL_FEDBILL ‐> DL_TERM_R

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Percent

DFEDALPHA ‐> DL_TERM_R

Week

Note: The figure plots impulse responses of repo financing to permanent liquidity injection shocks from a recursive VAR modelbased on the modified benchmark ordering. The black line is the median of the simulated responses, the blue line represents90% probability bands, and the green line represents 84% probability bands.

85

Figure 1.27: Response of Securities Out By Collateral Class to Permanent Open MarketOperation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

Week

Percent

DL_FEDTALL ‐> DLALL_OUT_US

0 5 10 15 20 25 30 35 40 45 50-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75 DL_FEDBILL ‐> DLALL_OUT_US

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

Percent

DFEDALPHA ‐> DLALL_OUT_US

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-0.5

0.0

0.5

1.0

1.5

2.0

2.5

Week

Percent

DL_FEDTALL ‐> DLALL_OUT_AGENCY

0 5 10 15 20 25 30 35 40 45 50-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25 DL_FEDBILL ‐> DLALL_OUT_AGENCY

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

DFEDALPHA ‐> DLALL_OUT_AGENCY

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Week

Percent

DL_FEDTALL ‐> DLALL_OUT_MBS

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0 DL_FEDBILL ‐> DLALL_OUT_MBS

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

DFEDALPHA ‐> DLALL_OUT_MBS

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Week

Percent

DL_FEDTALL ‐> DLALL_OUT_CORP

0 5 10 15 20 25 30 35 40 45 50-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00 DL_FEDBILL ‐> DLALL_OUT_CORP

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.5

-1.0

-0.5

0.0

0.5

1.0

Percent

DFEDALPHA ‐> DLALL_OUT_CORP

Week

Note: The figure plots impulse responses of collateralized financing to permanent liquidity injection shocks from a recursiveVAR model based on the modified benchmark ordering. The black line is the median of the simulated responses, the blue linerepresents 90% probability bands, and the green line represents 84% probability bands.

86

Figure 1.28: Response of Overnight and Continuing Securities Out By Collateral Class toPermanent Open Market Operation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

Week

Percent

DL_FEDTALL ‐> DLOUT_OC_US

0 5 10 15 20 25 30 35 40 45 50-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00 DL_FEDBILL ‐> DLOUT_OC_US

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

Percent

DFEDALPHA ‐> DLOUT_OC_US

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Week

Percent

DL_FEDTALL ‐> DLOUT_OC_AGENCY

0 5 10 15 20 25 30 35 40 45 50-1.5

-1.0

-0.5

0.0

0.5

1.0 DL_FEDBILL ‐> DLOUT_OC_AGENCY

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

DFEDALPHA ‐> DLOUT_OC_AGENCY

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Week

Percent

DL_FEDTALL ‐> DLOUT_OC_MBS

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25 DL_FEDBILL ‐> DLOUT_OC_MBS

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.75

-1.50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

DFEDALPHA ‐> DLOUT_OC_MBS

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

Week

Percent

DL_FEDTALL ‐> DLOUT_OC_CORP

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0 DL_FEDBILL ‐> DLOUT_OC_CORP

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

Percent

DFEDALPHA ‐> DLOUT_OC_CORP

Week

Note: The figure plots impulse responses of overnight and continuing collateralized financing by collateral class to permanentliquidity injection shocks from a recursive VAR model based on the modified benchmark ordering. The black line is the medianof the simulated responses, the blue line represents 90% probability bands, and the green line represents 84% probability bands.

87

Figure 1.29: Response of Term Securities Out By Collateral Class to Permanent Open MarketOperation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

Week

Percent

DL_FEDTALL ‐> DLOUT_TERM_US

0 5 10 15 20 25 30 35 40 45 50-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75 DL_FEDBILL ‐> DLOUT_TERM_US

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Percent

DFEDALPHA ‐> DLOUT_TERM_US

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

Week

Percent

DL_FEDTALL ‐> DLOUT_TERM_AGENCY

0 5 10 15 20 25 30 35 40 45 50-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0 DL_FEDBILL ‐> DLOUT_TERM_AGENCY

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-4

-3

-2

-1

0

1

Percent

DFEDALPHA ‐> DLOUT_TERM_AGENCY

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

Week

Percent

DL_FEDTALL ‐> DLOUT_TERM_MBS

0 5 10 15 20 25 30 35 40 45 50-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5 DL_FEDBILL ‐> DLOUT_TERM_MBS

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Percent

DFEDALPHA ‐> DLOUT_TERM_MBS

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

5

Week

Percent

DL_FEDTALL ‐> DLOUT_TERM_CORP

0 5 10 15 20 25 30 35 40 45 50-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5 DL_FEDBILL ‐> DLOUT_TERM_CORP

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-3

-2

-1

0

1

2

3

Percent

DFEDALPHA ‐> DLOUT_TERM_CORP

Week

Note: The figure plots impulse responses of term collateralized financing to permanent liquidity injection shocks by collateralclass from a recursive VAR model based on the modified benchmark ordering. The black line is the median of the simulatedresponses, the blue line represents 90% probability bands, and the green line represents 84% probability bands.

88

Figure 1.30: Response of Financing Fails By Collateral Class to Permanent Open MarketOperation Shocks


US

0 5 10 15 20 25 30 35 40 45 50-30

-20

-10

0

10

20

Week

Percent

DL_FEDTALL ‐> DLFAIL_US DL_FEDBILL ‐> DLFAIL_US

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-15

-10

-5

0

5

10

15

20

25

30

0 5 10 15 20 25 30 35 40 45 50-10

0

10

20

30

40

Percent

DFEDALPHA ‐> DLFAIL_US

Week

AG

EN

CY

0 5 10 15 20 25 30 35 40 45 50-15

-10

-5

0

5

10

15

20

Week

Percent

DL_FEDTALL ‐> DLFAIL_AGENCY DL_FEDBILL ‐> DLFAIL_AGENCY

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-10

-5

0

5

10

15

20

25

0 5 10 15 20 25 30 35 40 45 50-10

-5

0

5

10

15

20

25

30

Percent

DFEDALPHA ‐> DLFAIL_AGENCY

Week

MB

S

0 5 10 15 20 25 30 35 40 45 50-10

-5

0

5

10

15

20

25

30

Week

Percent

DL_FEDTALL ‐> DLFAIL_MBS DL_FEDBILL ‐> DLFAIL_MBS

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-5

0

5

10

15

20

25

30

35

0 5 10 15 20 25 30 35 40 45 50-5

0

5

10

15

20

25

30

Percent

DFEDALPHA ‐> DLFAIL_MBS

Week

CO

RP

0 5 10 15 20 25 30 35 40 45 50-6

-5

-4

-3

-2

-1

0

1

2

3

Week

Percent

DL_FEDTALL ‐> DLFAIL_CORP

0 5 10 15 20 25 30 35 40 45 50-6

-5

-4

-3

-2

-1

0

1

2

3 DL_FEDBILL ‐> DLFAIL_CORP

Percent

Week0 5 10 15 20 25 30 35 40 45 50

-5

-4

-3

-2

-1

0

1

2

3

Percent

DFEDALPHA ‐> DLFAIL_CORP

Week

Note: The figure plots impulse responses of financing fails by collateral class to permanent liquidity injection shocks from arecursive VAR model based on the modified benchmark ordering. The black line is the median of the simulated responses, theblue line represents 90% probability bands, and the green line represents 84% probability bands.

89

Chapter 2

Monetary Policy and the Non-bankFinancial Sector: A Look atCommercial Paper

I Introduction

In recent years, commercial paper has played a prominent role in credit activity. For example,

at the end of 2006, outstanding commercial paper peaked at close to $2 trillion dollars.1

Therefore, its economic importance is on par with the tri-party repo market in providing

short-term funding to the financial system.2

According to the Securities Industry and Financial Markets Association (SIFMA), com-

mercial paper makes up about 40% of all outstanding money market instruments. Moreover,

asset-backed paper represents nearly 60% of the total commercial paper market.3 These

institutional facts suggest that it is important to trace out the impact of monetary policy

through funding conditions in the commercial paper market in order to deeply understand

the transmission channels of monetary policy.

Funding conditions in the money markets operate as a feedback effect on the real econ-

1Federal Reserve Board Flow of Funds Release, table L.208. Accessed June 5, 2014.2Geithner (2008) reports that tri-party repos funded around $2.5 trillion of assets in early 2007.3According to Krishnamurthy et al. (2014), prior to the 2007-2008 financial crisis, asset-backed com-

mercial paper was the largest short-term liability issued by the shadow banking system. Bernanke (2010)describes shadow banks as being institutions that are not regulated depository institutions which includesasset-backed commercial paper conduits, money market mutual funds, and broker-dealer investment banks.

90

omy.4 When financial intermediaries expand their balance sheets, there is an increase in

the supply of credit and greater risk-taking. A balance sheet compression leads to pruden-

tial behavior: a reduction in liquidity, a flight to quality, a flight from maturity, and as a

by-product an increase in systemic risk.5

The largest originators of short-term credit to financial and corporate institutions are

money market mutual funds. Naturally, they are also the largest investors in the commercial

paper market. Taxable funds generally hold around one-third to 40% of all outstanding

paper, mainly for institutional investors. For example, at the start of January 2007, institu-

tional accounts represented 65% of the $2.18 trillion taxable money fund assets outstanding.

Retail accounts make up the rest.6

In this paper, we measure the dynamic structural effects of a monetary policy shock on

funding conditions in the commercial paper market. As the initial stages of the monetary

transmission mechanism generally take place through money markets, the actions of the

central bank can impact the market for commercial paper in systematically important ways.

For example, policy changes in the target interest rate can influence either the perception

of risk or the risk-tolerance among commercial paper investors.7 They can also affect the

pricing of assets which serve as collateral in asset-backed commercial paper programs.

Intermediaries in the financial system pool risk and help insure individuals against liq-

uidity risk.8 Money market mutual funds are financial intermediaries that accept deposits

and specialize in liquidity risk management. In order to attract deposits, a money fund

must maximize its portfolio return with respect to changes in the money supply and demand

for redemptions by fund investors. Along these lines, our work is motivated as a test for a

4It is widely documented that during the financial crisis of 2007-2008 runs on money market mutual fundsseverely disrupted the economic activities of commercial paper issuers. As a last resort, the Federal Reservecreated the Commercial Paper Funding Facility to “shield the allocation of real economic investment fromliquidity distortions created by the run.” (Adrian et al., 2011) For a historical narrative, see Financial CrisisInquiry Commission (2011).

5Here we define systemic risk as the propagation of default due to fire sales and or rollover risk.6Anderson and Garcon (2009).7Also known as the ‘risk-taking channel.’ See Borio and Zhu (2012) for a review.8See Diamond and Dybvig (1983).

91

liquidity risk channel of monetary transmission through the non-bank (“shadow”) banking

sector.

Consider a money fund which only invests in reserves and commercial paper, a money-

like substitute. Reserves do not pay interest but they do safeguard against demand for

fund redemptions. On the other hand, commercial paper pays a discount rate of interest

determined on the date of issue. When issued, the commercial paper rate is fixed and the

market price of the underlying claim becomes a function of capital market conditions. As

a result, investing in commercial paper allows the fund to pay dividends but exposes fund

assets to liquidation risk.

In the event of a policy tightening by the Federal Open Market Committee (FOMC), the

supply of reserves in the money market will decline. Thus, it will become more difficult to

access liquidity. This suggests that the money fund would need to hold more of its assets as

reserves in order to meet redemption demand by investors. However, any increase in reserve

holdings lowers the fund’s portfolio return. A lower return negatively impacts the ability to

attract deposits. It also affects the management fees earned by the fund.

Instead, the fund can engineer a liquid stream of reserve access and minimize the loss from

investment returns by shortening the maturity of its commercial paper holdings. Moreover,

the cost of early liquidation can be managed by investing in safe assets that can be easily

sold in order to raise cash should unanticipated redemptions occur. In other words, moving

towards medium-term paper with low credit risk will promote the ability to fund withdrawals

and still pay a competitive market yield. Funds would be unlikely to invest in risky paper

unless it has a short-term maturity. If a fund otherwise invests, it is more susceptible to

poor performance which could eventually lead to a suspension of funds as experienced by

BNP Paribas in August 2007 – an event which foreshadowed the coming financial crisis.

Of course, the willingness of money market mutual funds to invest in paper has important

implications for commercial paper issuers. Notably, commercial paper issuers generally repay

maturing commercial paper with newly issued paper. However, if money funds buy less paper

92

from risky borrowers, then many of these institutions will be unable to rollover their maturing

paper. As a result, raising funds in the commercial paper market may become problematic

if there is a policy tightening.

At the margin, issuers unable to directly place paper have two options – find alternative

financing or pay for immediacy. By paying for dealer assistance an issuer safeguards against

adverse funding conditions. Sponsoring broker-dealers intermediate credit on behalf of their

customers and purchase any paper unsold in the market. In this manner, the dealer’s will-

ingness to act as principal is a market signal about monitoring.9 As a result, borderline

issuers are no longer at the margin and liquidity is no longer a problem. In this manner,

changes in monetary policy are likely to affect the sources of funding among commercial pa-

per issuers. They also affect the cost of funding as issuers would need to substitute towards

dealer-intermediated paper.

Following much of the empirical literature on the monetary transmission mechanism, we

use a vector autoregression (VAR) to characterize the economic effects of a monetary policy

shock. However, the standard identification of a monetary policy shock in a VAR likely

suffers from endogeneity. One proposed solution is to combine high-frequency data with

a simple interest rate measure of market expectations.10 High-frequency data minimizes

measurement error by shrinking the observation window of information arrival. Market

interest rates control for omitted variables by summarizing all public information. Pursuing

this combined strategy, we estimate a VAR at the weekly frequency and include an indicator

of reserve imbalances as a forecast of expected changes in monetary policy the day before

the release of the FOMC’s policy statement. As part of our effort, we make the case that

our monetary policy shock is a qualified measure of unexpected changes in monetary policy.

Based on our high-frequency estimation, we find evidence of a strong liquidity effect. In

other words, a contractionary policy shock corresponds to an increase in the target federal

funds rate and a sharp decline in money market mutual fund assets. However, the behavior

9See Diamond (1984).10See Cochrane and Piazzesi (2002).

93

of institutional funds can vary drastically from retail funds.11 Therefore, we include the

assets of institutional and retail money funds in our analysis. We find institutional fund

assets sharply decline but assets at retail funds decline contemporaneously and then increase

after a long delay.

Consequently, contractionary monetary policy starts with a sharp reduction in available

liquidity in the money markets, led by a sharp reduction in assets at institutional money

funds. Yet the response of retail fund assets is less severe and as a result, suggestive of risk-

taking behavior. We next focus on the responses of commercial paper volumes in order to

trace the transmission of monetary policy to funding activity in the commercial paper market.

If a liquidity risk channel operates through the total supply of commercial paper, then

commercial paper volumes should behave consistently with our discussion of the transmission

mechanism.

For example, asset-backed conduits must routinely provide investors with credit guar-

antees. On the other hand, financial and nonfinancial issuers do not. Without guarantees,

investors would be unwilling to lend to commercial paper conduits. This implies that con-

duit issued paper, which is collateralized, is riskier than unsecured paper directly issued by

the conduit’s financial guarantor or a nonfinancial corporation with a comparable credit rat-

ing. For this reason, if a monetary contraction leads to diminished risk-taking by investors,

then we should see less asset-backed, more financial, and possibly more nonfinancial paper

activity.

Consistent with our rationale, we find that asset backed paper declines for nearly two

months after a contractionary shock. As evidence of a substitution towards less risky paper

holdings, financial paper sharply increases immediately following a contractionary monetary

policy shock. For the most part, nonfinancial does not respond. However, it does increase

two months after the shock. Similar to the decline in asset-backed paper, volumes for paper

with second tier credit ratings sharply decline for more than fifteen weeks. Thus, not only

11For example, Kacperczyk and Schnabl (2013) find that retail funds are much less sensitive to asset yielddifferentials and conclude retail funds do not engage in risk-taking behavior.

94

does monetary policy influence investors’ willingness to take on risk, the impact is stronger

for issuers with less liquid balance sheets.

Nor is this impact uniform across maturities. We also find evidence of a broad substitution

from long to short maturities at both the aggregate and issuer level. For example, volumes

of asset-backed, nonfinancial, and financial paper with maturities greater than 40 days all

show immediate declines.

Overall, these results support the existence of a liquidity risk channel for monetary pol-

icy operating through the total supply of commercial paper. This transmission mechanism

contributes to systemic risk in the real economy through aggregate rollover risk. An unex-

pected increase in the target rate leads to a reduction in aggregate money market mutual

fund assets. In turn, money funds ration credit to the riskiest borrowers. They also limit

their liquidity risk by lending at shorter maturities. The reduction in funding contributes to

rollover risk among issuers of commercial paper.

Furthermore, in the event a commercial paper conduit is unable to refinance maturing

paper, a default will propagate first to the broker-dealer subsidiary of its financial sponsor,

then to the sponsor, and then to the sponsor’s counterparties.12 The more interconnected

the sponsoring financial institution, the greater the resulting disruption in the supply of

credit to the real economy. As a result, we contend that the interplay between monetary

policy and commercial paper funding was a contributing factor to the panic of 2007-2008.13

This paper contributes to an emerging literature studying the transmission of monetary

policy to activity in the non-banking sector. While we are not the first to investigate a role

for commercial paper, we are the first to do so within a VAR framework with high-frequency

data. For example, Kashyap, Stein, and Wilcox (1993) use commercial paper volumes to

test for a bank loan supply channel of monetary policy at the quarterly frequency from the

Flow of Funds. By comparison, we aggregate daily information on outstanding commercial

12See Acharya et al. (2013).13This is one possible interpretation as to how, “the crisis spread from subprime housing assets to non-

subprime assets that have no direct connection to the housing market.” (Gorton and Metrick, 2012)

95

paper to the weekly frequency. In more recent work, Sunderam (2014) questions whether

the supply of asset-backed commercial paper is determined in response to it being perceived

as a low risk money substitute.

Presumably, other money market instruments also experience risk channel effects. For

instance, in the market for repurchase agreements, Boulware et al. (2014) observe that

unexpected changes in the federal funds rate lead to shorter repo market maturities. Conse-

quently, the results suggest that there is increase in rollover risk as a result of the tightening.

In comparison to Boulware et al. who focus on the volume of repo market activity in re-

sponse to policy shocks, Bech et al. (2012) examine the relationship between the federal

funds rate and the repo rate. In terms of unconventional policy instruments, D’Amico et

al. (2013) advance a repo scarcity channel operating through the total supply of Treasury

collateral.14

It is also possible that monetary policy can shape the portfolio risk borne by economic

agents. Adrian and Shin (2008) maintain that after a monetary tightening, deleveraging

by broker-dealer investment banks contribute to financial instability. Moreover, Chodorow-

Reich (2014) finds that asset purchase announcements by the FOMC do not lead to yield-

seeking behavior for financial institutions similarly subject to target rates of return.

The remainder of the paper is organized as follows. Section 2 outlines the institutional

details linking the stance of monetary policy, the market for commercial paper, and the

supply of credit. Section 3 describes the data and empirical methodology. Section 4 reports

the benchmark results after an unexpected monetary tightening. Section 5 discusses our

interpretation of the benchmark results and presents additional evidence including robustness

checks. Section 6 concludes.

14In related work, Cahill et al. (2013) finds the term structure of interest rates experience duration risksupply effects following FOMC Treasury purchase announcements.

96

II Institutional Background

Commercial paper represents an important segment of the fixed income market. On the one

hand, it provides issuers with cheap working capital. On the other, it can be used to deliver

income diversification for investors and lower risk since it has a relatively short maturity. In

this section, we discuss key institutional details regarding the supply of credit in the U.S.

commercial paper market. To begin, Figures 1 and 2 plot time series overlays of activity in

the market. Figure 1 shows the behavior of outstanding commercial paper volumes during

the sample, broken down by type of paper: asset-backed, financial, and nonfinancial. Figure

3 illustrates the amount of money market mutual fund assets over time.

II.1 The Market for Commercial Paper

Commercial paper is a short term discount note traded in the money markets. Therefore,

investors buy commercial paper at a time discount from an agreed upon future face value.15

Seasoned issuers rely on paper financing in order to roll over maturing paper continuously.

For issuers with strong credit ratings, commercial paper is a cheap substitute for bank loans.

Furthermore, because of its low regulatory costs, issuing commercial paper is a flexible

financing alternative. For example, paper must be originated with a maturity of less than

270 days. As it is not sold to the general public, it is exempt from SEC security registration

and maintenance.16 Consequently, it is traded in private markets.

Today, there are two groups of active issuers and three active markets. Corporate paper

is composed of nonfinancial and financial issuers and is directly issued as promissory by any

credit worthy institution. As we describe below, asset-backed paper is issued by a special

purpose vehicle known as the asset-backed commercial paper “conduit.”17

Figure 1 shows that during our sample (from 2001-2007), asset-backed paper grew in

15A typical face value is $100,000.16See the 1933 Securities Act, Section 3(a)(3).17Asset-backed conduits must also include credit enhancements and liquidity support. However, to increase

profits,some of these improvements were no longer being offered by sponsoring firms after 2003. (Andersonand Gascon, 2009)

97

close association with aggregate paper and financial paper. This observation reflects that

most conduits are sponsored by financial institutions involved in securitization. Conversely,

outstanding nonfinancial and financial paper show less substitutability. While both decline

from 2001 to 2003, there is very little correlation afterwards.

Though commercial paper is not sold to the general public, much of it is indirectly issued

to investors through money market mutual funds. As mentioned in the introduction, money

market mutual funds are the largest holders of commercial paper. Figure 2 plots aggregate,

institutional, and retail money market mutual fund assets over time.

II.2 Shadow Banking and the Supply of Credit

The stylized flow diagram in Figure 3 shows how commercial paper financing ultimately

extends to the real economy. The figure links monetary policy, commercial paper activity,

and the supply of credit to real economic investment.

The diagram starts with obligors, individuals and businesses seeking access to credit.

Examples include consumers applying for a car loan, firms leasing office space, and en-

trepreneurs paying startup costs with revolving credit. However, not all obligors are credit

constrained. In the case of large corporations, costly bank loans can be bypassed with cheap

commercial paper financing.

Loan originators are generally large bank holding companies but also include financial

service subsidiaries of nonfinancial companies such as the General Motors Acceptance Cor-

poration (GMAC) and General Electric Capital. After granting credit to obligors, some

originators finance a fraction of current assets by directly issuing paper. Alternatively, in a

“true sale,” loan originators sell loan obligations to a conduit that they may also sponsor.

Through the use of this special purpose vehicle, originators are able to lower their capital re-

quirements which promotes the ability of originators to extend credit. (Gorton and Souleles,

2007)

Conduits are passive bankruptcy remote special purpose vehicles. Their balance sheets

98

are composed of long-term loans repackaged into short term liabilities of commercial paper,

reflecting their role in maturity transformation. Like financial and nonfinancial paper, asset-

backed paper can be placed with investors directly but is overwhelmingly intermediated by

broker-dealer investment banks. Dealers finance market making duties by issuing financial

paper or by entering into a repurchase agreement.18

Whether commercial paper is placed directly or through a dealer, money market mutual

funds are the largest cash investors in the commercial paper market. Money funds accept

deposits from institutional and retail investors while pledging to preserve each dollar de-

posited. As recounted by Krishnamurthy et al. (2012), each dollar deposited represents

the origination of short term funding to the shadow banking system and subsequently, real

economic investment.

The final agent in Figure 3 is the monetary authority. On a daily basis, trades occur

between the Federal Reserve Bank of New York Open Market Desk and the Primary Gov-

ernment Securities Dealers of the Federal Reserve System (primary dealers). These trades

are open market operations.

An open market operation can be either permanent or temporary. For example, in a

temporary liquidity injection, the open market desk lends short term funds to a primary

dealer. That is, the interaction between a primary dealer and the open market desk is

analogous to a primary dealer selling financial paper to a money market mutual fund.

Furthermore, many of these dealers are subsidiaries of bank holding companies originating

loans, issuing financial paper, and sponsoring asset-backed commercial paper conduits.19 In

turn, open market operations can affect the amount of credit that the dealers and their parent

bank holding companies can ultimately issue. In this manner, monetary policy directly

transmits to real activity.

18Copeland and Martin (2012) present a daily breakdown of the collateral backing overnight tri-partyrepurchase agreements. Furthermore, dealers can extend credit to finance purchases of commercial paper forother market participants through reverse repurchase agreements.

19A current list of the primary dealers is maintained by the Federal Reserve Bank of New York at:http://www.newyorkfed.org/markets/pridealers_current.html.

99

http://www.newyorkfed.org/markets/pridealers_current.html


Measuring the response of commercial paper activity to an exogenous monetary policy shock

requires the careful isolation of monetary policy shocks.20 In the following section, we char-

acterize our empirical model of the monetary transmission mechanism. First, we introduce

and summarize the main variables used in our study. Lastly, we present and discuss our

identification assumptions.

III.1 Data

Table 1 presents summary statistics for all of the variables used in our analysis including

simple measures of dispersion and persistence. Our sample ranges from the week of January

5, 2001 to February 2, 2007. We end our sample the week before the first bank failure prior

to the crisis.

Our data on commercial paper is published by the Federal Reserve Board of Governors

on its website. It is aggregated from daily transactions cleared and settled by the Depository

Trust & Clearing Corporation (DTCC). All of the commercial paper data we use is based

on the April 10, 2006 revision.

The effective federal funds rate is collected from the Board of Governors H.15 release and

the federal funds target rate is from Board of Governors press releases. Meeting dates of the

Federal Open Market Committee are published on the Board of Governors website.

Money market mutual fund data is published in the Board’s H.6 release. Institutional and

retail money fund assets are constructed from data collected by the Investment Company

Institute (ICI). In addition, assets at retail money funds exclude retirement account balances.

We also include a measure of labor market conditions and energy prices. In particular,

we use the futures price of crude oil for the second delivery month as released by the U.S.

Department of Energy in its weekly Petroleum Status Report. Daily futures prices are the

20See Bernanke and Blinder (1992) and Christiano et al. (1999).

100

official daily closing prices from the trading floor of the New York Mercantile Exchange

(NYMEX) from Reuters, Inc.

The four week average of initial jobless claims is from the Unemployment Insurance

Weekly Claims Report released by U.S. Department of Labor Employment & Training Ad-

ministration. Initial claims are compiled weekly and show the number of individuals filing

for unemployment insurance for the first time.

All variables except interest rates have been transformed into growth rates.21 Interest

rate measures are in first differences except for deviations in the overnight rate from the

target which is in levels. Detailed descriptions of our data and its sources are listed in the

Data Appendix.

III.2 Modeling the Federal Open Market Committee

Our choice for studying the impact of monetary policy on economic variables is to use an

identification scheme which decomposes the response of the central bank into endogenous

and exogenous variation. Following the literature, we identify a monetary policy shock with

the residuals from the following ordinary least squares equation:

pw = Ψ(ΩFOMCw

)+ λpεp

w (2.1)

In other words, the policy instrument (pw) is equal to a linear combination of the cur-

rent economic state observed by the FOMC when setting the policy instrument(ΩFOMCw

)each week and a positive serially uncorrelated shock (λpεp

w) orthogonal to the FOMC’s ob-

served state of the economy as a result of contemporaneous information lags. Using the

model’s reduced form representation, we can measure the dynamic responses of a variable

to a monetary policy shock by estimating the following VAR:

Ψw = A(L)Ψw−1 + uw (2.2)

21Each variable was tested for a unit root using the augmented Dickey-Fuller test following Enders (2009).

101

where A(L) ≡ I −A1L is defined as the auto-regressive lag polynomial of order one, and

uw is a vector of reduced form residuals.22 After rearranging the VAR, we equate uw to the

structural economic shocks εw as:

uw = Λεw (2.3)

Identification of the underlying structural monetary policy shock, εpw, requires a set of

restrictions to be imposed upon Λ. We elaborate on ours in the following sub-section.

III.3 Identification

Our aim is to present evidence on the transmission of monetary policy to commercial paper

activity. Identification requires the selection of a limited vector of variables, Ψw, observed

each week w. In addition to a direct measure of Federal Reserve policy, variables measuring

labor market conditions and inflation are included in order to capture the systematic portion

of the Fed’s reaction function as dictated by its dual mandate.

Thus, our benchmark VAR is comprised of the target federal funds rate, the four week

moving average of jobless claims, and the futures price of crude oil. We also include the

assets of money market mutual funds to control for the effects of changes in money market

demand along with a measure of commercial paper activity. Lastly, we introduce a measure

of anticipated policy actions, MISS, the gap between the effective federal funds rate and the

target rate the day before a scheduled FOMC policy statement.23

One criticism with identifying monetary policy shocks by estimating the FOMC’s policy

rule in a VAR is that the VAR methodology restricts the number of variables that can be

included in the FOMC’s information set.24 Consequently, policy shocks identified by the

VAR as being unanticipated may instead have been fully anticipated. By using a market

22For simplicity, we only consider a first order VAR and exclude deterministic regressors.23The daily deviation in the effective rate from its target is a realized exogenous shift in the supply of

reserves due to a forecast miss by the FRBNY open market desk. For institutional details, see Carpenterand Demiralp (2006) and Hamilton (1997).

24An alternative is to augment the VAR with a factor model. See Bernanke and Boivin (2005).

102

measure of anticipated policy actions (through MISS), we can reflect all publicly available

information including changes in regime in the FOMC’s information set.25

To test statistically whether the MISS has predictive power for our multivariate system

we conduct a block exclusion test on thirteen lags of the MISS. The null hypothesis is

that the lags of the MISS can be omitted from each equation in the VAR. The likelihood

ratio test has 65 degrees of freedom and the null hypothesis is rejected at the one percent

level.26 Therefore, the lagged state of anticipated policy actions is not block exogenous and

strongly predicts the variables included in the VAR system. In summary, our benchmark

VAR includes the following six variables:

ΨTw = [OILw, MISSw, TARGETw, CPw, CLAIMSw, MMMFw] (2.4)

Moreover, the relationship between our reduced form residuals and structural distur-

bances defined by equation (3) can now be expressed as:

u1w

u2w

u3w

u4w

u5w

u6w

=

b11 b12 b13 b14 b15 b16

b21 b22 b23 b24 b25 b26

b31 b32 b33 b34 b35 b36

b41 b42 b43 b44 b45 b46

b51 b52 b53 b54 b55 b56

b61 b62 b63 b64 b65 b66

εOILw

εMISSw

εTARGETw

εCPw

εCLAIMSw

εMMMFw

(2.5)

Each structural disturbance is serially uncorrelated and has a covariance matrix equal to

the identity matrix. If we replace E[uwu

Tw

]= Σu by its sample analogue, Σu has n(n+1)

2= 21

free parameters and the Λ matrix contains 36 elements. Therefore, n(n−1)2

= 15 additional

restrictions are necessary and sufficient to estimate an exactly identified system.

Information Arrival

Orthogonalization constitutes the final restrictions needed for identification. Depending on

the size of the observation window, the inclusion of variables linked to financial markets

25This is the definition of the efficient markets model given by Fama (1970).26We also tested lags of the MISS for block exclusion with alternate measures of OIL and CLAIMS and

find similar results.

103

makes it likely that policy shocks from the VAR suffer from endogeneity.27 With high

frequency data, one can shrink the observation window and minimize their contamination.

At the weekly frequency we construct timing restrictions that mimic the daily release of

economic and financial market information. Moreover, because the MISS is observed the day

before a scheduled policy announcement, we are able to measure market expectations at the

end of that particular day.28 To illustrate our point, Figure 5 shows a timeline of information

arrival for each variable in our VAR for the week ending Friday July 2, 2004. Following this

example, we adopt a Choleski decomposition with the same recursive structure:

u1w

u2w

u3w

u4w

u5w

u6w

=

b11 0 0 0 0 0b21 b22 0 0 0 0b31 b32 b33 0 0 0b41 b42 b43 b44 0 0b51 b52 b53 b54 b55 0b61 b62 b63 b64 b65 b66

εOILw

εMISSw

εTARGETw

εCPw

εCLAIMSw

εMMMFw

(2.6)

In other words, the restricted Λ matrix is equivalent to constructing policy shocks where

news about the economy on the day of a FOMC policy statement does not affect the policy

decision. Therefore, our identified policy shocks are orthogonal to macroeconomic news

released within the event window.29

Unless stated otherwise, the estimated impulse response functions noted in the rest of

the paper are based on the above benchmark identification. Our VAR includes thirteen

weeks of lagged variables and a constant. We also added a dummy variable for the weeks of

the September 11, 2001 terrorist attacks and for the 53rd week of the year. Following Sims

and Zha (1999) we report median, 68%, and 90% probability intervals for simulated impulse

responses.

27For example, Rigobon and Sack (2003) measure the response of monetary policy to daily movements inthe stock market.

28See Furfine (1999) for the timing of intraday payments in the federal funds market.29A summary of the event study identification approach is outlined by Gurkaynak and Wright (2013).

104

The Federal Open Market Committee Reaction Function

Our assumption about the FOMC reacting systematically to observable macroeconomic con-

ditions implies that the TARGET should respond in a theoretically consistent manner to

unexpected changes in OIL and CLAIMS. To test this hypothesis, we shock OIL and CLAIMS

separately and plot the TARGET response. This exercise is analogous to estimating policy

reaction functions in the form of equation (1).30

Figure 6 displays the impulse response function of TARGET after a shock to OIL in the

top panel and a shock to CLAIMS in the bottom panel. As expected, an OIL shock leads

to an increase in the target rate contemporaneously, with the peak effect coming after 10-12

weeks. Following a shock to claims, the funds rate declines. After a two week delay the

response is persistent and still falling after 52 weeks.

Furthermore, a one standard deviation shock to OIL is close to 4 times bigger than a one

standard deviation shock to CLAIMS. This suggests that during our sample the FOMC was

much more sensitive to conditions in the labor market than inflationary pressures emanating

from the energy market. Notably, the FOMC lowered the federal funds rate from 6.5% in

January 2001 to 1% in June 2003 where it remained for one year.31 Thus, our estimates for

the FOMC reaction function behave in a plausible manner.

IV Results

In the following section we present impulse responses of commercial paper activity following

an unexpected increase in the federal funds target interest rate. Our estimated VAR is based

on the benchmark identification outlined above.

The first section displays the responses of macroeconomic targets and indicators of mone-

tary policy. The second section compares the responses of money market mutual fund assets,

key investors in the commercial paper market and an indicator of aggregate liquidity in the

30See Bernanke and Blinder (1992).31Until October 2008, 1% was a historic low.

105

money market. Lastly, we present the responses for outstanding commercial paper at the

aggregate level and by type of issuer.

IV.1 Jobless Claims, Oil, and the Interest Rate Response

In order to build confidence in our monetary policy shocks, we first display the dynamic

response of our macroeconomic target variables and market interest rates. Figure 7 shows

the responses of CLAIMS, OIL, MISS, and TARGET to a positive one standard deviation

in the federal funds rate target.

The top two panels show the four week moving average of initial jobless claims and

the spot price of oil do not respond to a TARGET shock. However, these estimates are

consistent with economic theory. For example, OIL does not suffer from a price puzzle, a

standard finding in recursively identified monetary VARs.32

Furthermore, the insignificant response of CLAIMS is possibly due to our use of high

frequency data which is unlikely to be related to changing macroeconomic fundamentals in

the same way as monthly and quarterly data. It may also reflect that policy had a limited

impact on labor market activity during our sample.

The response of TARGET to its own shock is plotted in the bottom left panel. The

target rate increases contemporaneously around 6 basis points, and after a sharp increase in

week 5 increases to 8 basis points after one year. Consistent with FOMC policy, an increase

in the target rate is permanent one year later. Furthermore, after the initial unexpected

shock other than a small increase in the 5th week a tightening cycle initiated by the FOMC

is fully anticipated.

32For a recent discussion of recursively identified monetary VARs see Barakchian et al. (2013).

106

If a shock to TARGET is truly unanticipated then the MISS response should be insignifi-

cant. Other than marginal significance in the first 8 weeks there is no evidence of anticipatory

behavior in response to our policy shocks. Furthermore, the increase week 7 suggests the

TARGET include shocks to forward guidance.

IV.2 Money Market Mutual Fund Responses

Figure 8 plots the response of assets held by money market mutual funds. For each variable,

we ran a separate VAR. The middle panel displays the response of institutional money funds

and the bottom panel plots retail assets. Money funds are important indicators of liquidity

in the commercial paper market and their response is key in our sketch of the liquidity risk

channel.

All three variables show declines, however, the decline of retail money funds is only

contemporaneous. In fact, by week 22, retail assets are positive and significant. On the

other hand, institutional money funds face persistent redemptions going out to a year. Con-

sequently, a monetary tightening leads to a reduction in liquidity in the money market

dominated by institutional investors. The TARGET’s impact on retail funds is smaller. As

a result retail funds see their assets increase over time, possibly creating an incentive to take

on risk.

IV.3 Commercial Paper Responses

The response of money fund assets shown in the previous section suggests there is a reduction

in aggregate liquidity in the money markets after a shock to TARGET. If money funds face

an increase in demand for liquidity by investors then their investment choices should also be

affected. In Figure 9, we plot the responses of four measures of commercial paper activity:

aggregate, financial, nonfinancial, and asset-backed paper outstanding. Each response was

estimated by running our benchmark VAR using each variable individually.

As seen in the top left panel after a shock to TARGET outstanding paper increases

107

persistently after one quarter up to 41 weeks. However, the response of outstanding paper

by issuer supports our hypothesis of money funds behaving more prudently. For example,

financial paper increases contemporaneously lasting four weeks. Nonfinancial paper declines

briefly in week 1 but then shows an increase peaking in week 18. Lastly, asset backed paper

shows the least significant response but the point estimate shows a decline in week 2.

Combined with the responses of money funds, it is clear that a shock to TARGET leads

to a shift in the allocation of lending to issuers of commercial paper. Furthermore, this shift

depends on the type of issuer. In addition, we see evidence that the ability of financial firms

to obtain credit relative to a conduit matters. Moreover, nonfinancial paper had the largest

response suggesting commercial paper financing is dedicated funding whereas financial firms

are primarily using this market as a temporary substitute.

V Discussion

In the previous section we showed evidence supporting a liquidity risk channel for monetary

policy. However, these results alone are not convincing when using aggregate data. In the

following section we explore in more detail whether our liquidity risk hypothesis is supported

by the data.

In the first section, we compare the response of paper issued directly versus through a

conduit. The second section explores whether our benchmark results are driven purely by

credit risk. Next we explore whether intermediation programs matter for investors. In the

fourth section we ask whether our basic findings are different when controlling for the length

of the financing agreement. We conclude with a brief summary of the various robustness

checks conducted.

108

V.1 Responses by Collateralization

In the previous section we found evidence of a financing relationship between financial in-

stitutions and their conduits. In the following section we explore whether this relationship

is a result of direct issuance by the company or indirect through a conduit. Figure 10 plots

the responses of asset backed paper and unsecured paper which is the sum of financial and

nonfinancial paper.

When we compare the responses by column, unsecured paper increases and secured paper

decreases. When comparing the responses by rows, the magnitudes are the same but the

significance depends on whether we include retail or institutional fund assets. For asset

backed paper, including retail money fund assets leads to a persistent decline lasting 8 weeks.

In contrast, the response of financial paper is more persistent when including institutional

money fund assets in the VAR. These results support the idea that a shock to the TARGET

leads to retail funds reducing exposure in paper indirectly issued and institutional funds

increase exposure in paper issued directly.

V.2 Responses by Credit Rating

Up till now we have shown evidence supporting a liquidity risk channel of monetary policy.

However, our results could also just be driven by a shift away from second tier to first tier

rated paper. In figure 11 we plot the response of these two variables to a shock in TARGET.

Other than for the first week the response of tier-1 paper is insignificant. For tier-2

paper, however, there is a persistent decline peaking after one quarter. Furthermore, the fall

in CPT2 is much larger. The contrast in responses suggests investors are discriminatory in

their willingness to take on risk after a TARGET shock.

109

For example we would expect the response of tier-1 paper to be positive and significant.

However, overall it is insignificant and the only week of significance is where the point

estimate is negative. Combined with evidence on the responses in the previous section, it

would seem investors are most willing to lend to issuers that have the ability to find alternate

financing when maturing paper comes due.

V.3 Responses by Intermediation

As we described in the introduction, a reduction in money market liquidity should also

impact the relative cost of intermediation programs for commercial paper issuers. In figure

12 we plot the response of directly placed and dealer placed paper after a shock to TARGET.

There is no significant response for paper placed directly and a marginal negative response

for dealer placed paper peaking in week 5 at -0.5%. Based on these results we conclude the

method of intermediation alone is not a significant factor in determining investors’ willingness

to lend to issuers of commercial paper.

V.4 Responses by Paper Maturity

Although we have shown evidence that suggests that monetary policy contributes to rollover

risk in the market for commercial paper, credit shocks may also impact the maturity structure

of lending. For example, for prudent investors, it is possible that lending to a risky borrower

overnight is no different that lending to a borrower of investment grade for 30 days.

We think this question is important as many observers of the 2007-2008 crisis suggest that

adverse money market conditions contributed to the shift towards short-term financing.33

Moreover, policymakers still debate the concerns that they have about rollover risk among

institutions with significant maturity mismatch.

In Figures 13 to 16 we plot the responses of commercial paper issuance by maturity.

Each response is estimated from a VAR with the following ordering: OIL, MISS, TARGET,

33Brunnermier (2009) is one of the most well known to make this claim.

110

CLAIMS, IMMMF, and CP. While each variable has six maturities there is a clear pattern

for all of them so we only show the strongest four responses.

Looking at the four figures, a pattern emerges supporting a maturity shift from long term

financing to short term financing after a TARGET shock. For example, for the aggregate

level, there is an immediate decline in paper issued with maturities greater than 40 days.

For maturities between 10 and 40 days, there is an increase with paper maturing in between

10 and 20 days showing persistence out to 52 weeks.

When we look at the responses by issuer type, we see a similar pattern for paper with the

longest maturities. For example, volumes of asset-backed, nonfinancial, and financial paper

with maturities greater than 40 days all show immediate declines. In contrast, the shorter

maturities show increases where the timing seems to be once again related to the type of

issuer and hence their balance sheet strength. Volumes of asset-backed paper maturing in 1

to 4 and 21 to 40 days, nonfinancial in 5 to 20 days, and financial in 10 to 40 days increase.

In conclusion, we argue the results indicate that monetary policy contributes to liquidity

risk in the money market which leads to rollover risk for issuers of commercial paper. Fur-

thermore, the impact of the transmission channel is related to the perceived balance sheet

strength of each issuer. For example, commercial paper conduits have large maturity gaps

on their balance sheet. Therefore, they are more susceptible to default if they are unable to

rollover their maturing paper because they have limited alternative financing options.

V.5 Robustness

We also estimated the benchmark VAR with different orderings and lag lengths. Moreover,

it is possible our shocks could be different if we include different macroeconomic targets. We

therefore used the spot price of crude oil and futures oil price from the first delivery month

to see if our results changed. Furthermore, we use other labor market variables included in

the weekly claims report such as continuing claims and the insured unemployment rate. In

all cases we found comparable results to the ones we presented.

111

VI Conclusion

The lessons from the 2007-2008 financial crisis emphasize a need to understand the macroe-

conomic importance of funding market conditions impact on the supply of credit to real

economic investment. However, one question left unexplained is the role of monetary policy.

The purpose of this paper is to answer whether monetary policy transmits to the com-

mercial paper market by shaping investors’ willingness to take on risk. In addition to being

a contributing factor in the 2007-2008 financial crisis, the market for commercial paper is

one of the largest dedicated funding markets in the world.

The results presented here support a liquidity risk channel for monetary policy operating

through the total supply of credit to the commercial paper market. We find an unexpected

increase in the target federal funds rate leads to a persistent increase in money market mutual

fund redemptions and by extension, a reduction in money market liquidity. As a by-product,

lenders ration credit to the riskiest borrowers. Furthermore, we show this liquidity risk effect

is a function of not only perceived borrower credit risk but also the duration of this risk.

Subsequently, we infer the transmission of monetary policy contributes to systemic risk

in the macroeconomy through refinancing risk. Moreover, we show the FOMC can influence

the supply of credit to money like substitutes such as commercial paper causing unintended

consequences. Therefore, commercial paper activity should be a determining consideration

in macroprudential regulation.

112

References

Acharya, Viral V., Philipp Schnabl, and Gustavo Suarez. 2010. Securitization without risktransfer. Journal of Financial Economics, 107(3): 515-535.

Adrian, Tobias, Karin Kimbrough, and Dina Marchioni. 2011. The federal reserve’s com-mercial paper funding facility. FRBNY Economic Policy Review, May: 25-39.

Adrian, Tobias, and Hyun Song Shin. 2008. Financial intermediaries, financial stability andmonetary policy. Paper presented at the Federal Reserve Bank of Kansas City Sympo-sium at Jackson Hole, Wyoming, August 21-23.

Anderson, Richard G., and Charles S. Gascon. 2009. The commercial paper market, thefed, and the 2007-2009 financial crisis. Federal Reserve Bank of St. Louis Review, Novem-ber/December, 91.

Altunbasa, Yener, Leonardo Gambacortab, and David Marques-Ibanez. 2014. Does Mone-tary Policy Affect Bank Risk? International Journal of Central Banking, 10(1): 95-135.

Barakchian, S. Mahdi, and Christopher Crowe. 2013. Monetary policy matters: evidencefrom new shocks data. Journal of Monetary Economics, 60(8): 950-966.

Bech, Morten, Elizabeth Klee, and Victor Stebunovs. 2012. Arbitrage, liquidity and exit:the repo and federal funds markets before, during, and emerging from the financial crisis.FRB Finance and Economics Discussion Series, 2012-21: 1-54.

Bernanke, Ben S.. 2010. Statement before the financial crisis inquiry commission. 2 Septem-ber.

Bernanke, Ben S., and Alan S. Blinder. 1992. The federal funds rate and the channels ofmonetary transmission. American Economic Review, 82(4): 901-921.

Bernanke, Ben S., Jean Boivin, and Piotr Eliasz. 2005. Measuring the effects of mone-tary policy: a factor-augmented vector autoregressive (FAVAR) approach. The QuarterlyJournal of Economics, 120(1): 387-422.

113

Borio, Claudio, and Zhu Haibin. 2012. Capital regulation, risk-taking and monetary policy:a missing link in the transmission mechanism? Journal of Financial Stability, 8(4): 236-251.

Boulware, Karl David, Jun Ma, and Robert Reed. 2014. How do money market conditionsaffect shadow banking activity? evidence from security repurchase agreements. Mimeo.

Brunnermeier, Markus K.. 2009. Deciphering the liquidity and credit crunch 2007-2009.Journal of Economic Perspectives, 23(1): 77-100.

Cahill, Michael E., Stefania D’Amico, Canlin Li, and John S. Sears. 2013. Duration riskversus local supply channel in treasury yields: evidence from the federal reserve’s assetpurchase announcements. FRB Finance and Economics Discussion Series, 2013-35.

Calomiris, Charles W., Charles P. Himmelberg, and Paul Wachtel. 1995. Commercial pa-per, corporate finance, and the business cycle: a microeconomic perspective. Carnegie-Rochester Conference Series on Public Policy, 42: 203-250.

Carpenter, Seth, and Selva Demiralp. 2006. The liquidity effect in the federal funds mar-ket: evidence from daily open market operations. Journal of Money, Credit and Banking,38(4): 901-920.

Chodorow-Reich, Gabriel. 2014. Effects of Unconventional Monetary Policy on FinancialInstitutions. NBER Working Paper Series, 20230.

Cochrane, John H., and Monika Piazzesi. 2002. The fed and interest rates - a high-frequencyidentification. American Economic Review, 92(2): 90-95.

Copeland, Adam, Antoine Martin, and Michael Walker. 2012a. Repo runs: evidence fromthe tri-party repo market. Federal Reserve Bank of New York Staff Reports, No. 506.

Christiano, Lawrence J., Martin Eichenbaum, Charles L. Evans. 1999. Monetary policyshocks: what have we learned and to what end? In: Taylor, J.B., Woodford, M. (eds.),Handbook of Macroeconomics. North-Holland, Amsterdam.

D’Amico, Stefania, Roger Fan, and Yuriy Kitsul. 2013. The scarcity value of Treasurycollateral: repo market effects of security-specific supply and demand factors. FederalReserve Bank of Chicago Working Paper, 2013-22.

Diamond, Douglas W.. 1984. Financial intermediation and delegated monitoring. The Re-view of Economic Studies, 51(3): 393-414.

Diamond, Douglas W., and Philip H. Dybvig. Bank runs, deposit insurance, and liquidity.Journal of Political Economy, 91(3): 401-419.

114

Disyatat, Piti. 2011. The bank lending channel revisited. Journal of Money, Credit andBanking, 43(4): 711-734.

Enders, Walter. 2009. Applied econometric time series. John Wiley & Sons.

English, William B., Skander J. Van den Heuvel, and Egon Zakrajsek. 2013. Interest raterisk and bank equity valuations. FRB Finance and Economics Discussion Series, 2012-26.

Fama, Eugene F. 1970. Efficient capital markets: a review of theory and empirical work.The Journal of Finance, 25(2): 383-417.

Farhi, Emmanuel, and Jean Tirole. 2009. Leverage and the central banker’s put. The Amer-ican Economic Review, 99(2): 589-593.

Financial Crisis Inquiry Commission. 2011. The financial crisis inquiry report. New York:Public Affairs.

Friedman, Benjamin M., and Kenneth N. Kuttner. 2011. Implementation of monetary pol-icy: how do central banks set interest rates? In: Friedman and Woodford (eds.), Handbookof Monetary Economics. Amsterdam, North-Holland.

Furfine, Craig H. 1999. The microstructure of the federal funds market. Financial Markets,Institutions & Instruments, 8(5): 24-44.

Gorton, Gary, and Andrew Metric. 2010. Securitized banking and the run on repo. Journalof Financial Economics, 104(3): 425-451.

Geithner, Timothy F.. 2008. Reducing system risk in a dynamic financial system. Speechbefore the Economic Club of New York, 9 June.

Gorton, Gary B., and Nicholas S. Souleles. 2007. Special purpose vehicles and securitization.In The risks of financial institutions, University of Chicago Press, 549-602.

Gurkaynak, Refet S., and Jonathan H. Wright. Identification and inference using event stud-ies. The Manchester School, 81(S1): 48-65.

Hamilton, James D.. 1997. Measuring the liquidity effect. American Economic Review,87(1): 87-97.

Jimenez, Gabriel, Gabriel, Steven Ongena, Jose Luis Peydro, and Jesus Saurina. 2014. Haz-ardous times for monetary policy: what do twenty three million bank loans say about theeffects of monetary policy on credit risk taking?” Econometrica 82(2): 463-505.

Kacperczyk, Marcin, and Philipp Schnabl. 2010. When safe proved risky: commercial paperduring the financial crisis of 2007-2009. Journal of Economic Perspectives, 24(1): 29-50.

115

——. 2013. How safe are money market funds? The Quarterly Journal of Economics, 128:1073-1122.

Kashyap, Anil K., Jeremy C. Stein, and David W. Wilcox. 1993. Monetary policy and creditconditions: evidence from the composition of external finance. American Economic Re-view, 83(1): 78-98.

Krishnamirthy, Arvind, Stefan Nagel, and Dmitry Orlov. 2014. Sizing up repo, Journal ofFinance, forthcoming.

Morris, Stephen, Hyun Song Shin. 2014. Risk-taking channel of monetary policy: a globalgame approach. Mimeo.

Reed, Robert R., and Ejindu S. Ume. 2014. Housing and monetary policy. Mimeo.

Rigobon, Robert, and Brian Sack. 2003. Measuring the reaction of monetary policy to thestock market. Quarterly Journal of Economics, 118(2): 639-669.

Sims, Christopher A., and Tao Zha. 1999. Error bands for impulse responses. Econometrica,67(5): 1113-1156.

Sunderam, Adi. 2013. Money creation and the shadow banking system. Mimeo.

116

Data Appendix

This appendix documents the datasets used in our empirical analysis and is organized by

observation frequency.

Data By Meeting

FOMC Meeting Calendar - The meeting calander of the FOMC is published online by the

Board of Governors at: http://www.federalreserve.gov/monetarypolicy/fomccalendars.

htm.

Interest Rates - The H.15 release is published online by the Board of Governors at: http://

www.federalreserve.gov/releases/h15. The daily effective federal funds rate is the daily

weighted average rate on brokered trades at the 6:30 p.m. close of Fedwire. Historical changes

of the target federal funds rate is published online by the Federal Reserve Bank of New York

at: http://www.newyorkfed.org/markets/statistics/dlyrates/fedrate.html.

Data By Week

Commercial Paper - Commercial paper outstanding and issuance is published online by

the Board of Governors at: http://www.federalreserve.gov/releases/cp/default.htm.

Outstanding paper is released Wednesday at the close of business. Issued paper is released

daily at 9:45 a.m..

Energy Prices - The Weekly Petroleum Status Report is published online by the U.S.

Department of Energy Information Administration at: http://www.eia.gov/petroleum/

117

http://www.federalreserve.gov/monetarypolicy/fomccalendars.htm

http://www.federalreserve.gov/monetarypolicy/fomccalendars.htm

http://www.federalreserve.gov/releases/h15


http://www.newyorkfed.org/markets/statistics/dlyrates/fedrate.html

http://www.federalreserve.gov/releases/cp/default.htm

http://www.eia.gov/petroleum/supply/weekly/


supply/weekly/. Futures price for the second delivery month is the official daily closing

price at 2:30 p.m. from the New York Mercantile Exchange (NYMEX).

Jobless Claims - The Unemployment Insurance Weekly Claims Report is published online

by the U.S. Department of Labor Employment and Training Administration at: http:

//www.oui.doleta.gov/unemploy/data.asp. The 4-week moving average is released every

Thursday at 8:30 a.m..

Money Market Mutual Fund Balances - The H.6 release is published online by the

Board of Governors at : http://www.federalreserve.gov/releases/h6. Assets at money

funds is released every Thursday at 4:30 p.m..

Seasonal Adjustment - Variables which have not been adjusted for seasonality are first

estimated using a seasonal dummy model.

118



http://www.oui.doleta.gov/unemploy/data.asp

http://www.oui.doleta.gov/unemploy/data.asp


Table 2.1: Descriptive Statistics: January 5, 2001 to February 2, 2007Obs. Max Min µ σ ρ

Panel A: Energy PricesWest Texas Intermediate Crude Oil Futures Price (OILw) 315 10.01 -26.23 0.194 4.73 0.993

Panel B: Cost of CreditFederal Funds Target Rate (TARGETw) 315 25.00 -50.00 0.060 8.24 0.995Gap Between the Overnight and Target Rate (MISSw) 315 27.00 -25.00 0.493 4.21 -0.018

Panel C: Commercial Paper by IssuerCommercial Paper Outstanding (CPw) 315 4.80 -3.71 -0.062 1.55 1.009Unsecured Commercial Paper Outstanding (UCPw) 315 6.02 -5.07 -0.033 2.11 0.993Asset Backed Commercial Paper Outstanding (ABCPw) 315 3.33 -2.07 -0.125 1.64 1.005Financial Commercial Paper Outstanding (FINCPw) 315 6.77 -4.41 -0.031 2.50 0.999Nonfinancial Commercial Paper Outstanding (NONFINCPw) 315 7.73 -8.06 -0.037 3.66 0.985

Panel D: Commercial Paper by Credit RatingTier-1 Rated Commercial Paper Outstanding (CPT1w) 315 3.87 -3.34 0.001 0.78 1.007Tier-2 Rated Commercial Paper Outstanding (CPT2w) 315 30.54 -9.92 -0.003 3.39 0.959

Panel E: Commercial Paper by IntermediationDirectly Placed Commercial Paper Outstanding (DIRPCPw) 315 8.78 -6.61 0.005 2.07 0.993Dealer Placed Commercial Paper Outstanding (DPCPw) 315 5.24 -6.22 -0.001 1.17 1.004

Panel F: Real ActivityFour Week Moving Average of Initial Jobless Claims (CLAIMSw) 315 7.75 -5.03 -0.005 1.58 0.992

Panel G: Money Market Mutual Fund AssetsMoney Market Mutual Fund Balances (MMMFw) 315 4.61 -1.71 0.048 0.70 0.995Institutional Money Market Mutual Fund Balances (IMMMFw) 315 7.34 -2.26 0.110 0.99 0.974Retail Money Market Mutual Fund Balances (RMMMFw) 315 1.29 -1.78 -0.052 0.49 0.997

Panel H: Commercial Paper by MaturityIssued CP maturing in 10 to 20 days (CP10TO20w) 315 79.36 -88.27 -0.204 28.18 0.373Issued CP maturing in 21 to 40 days (CP21TO40w) 315 90.31 -120.83 -0.120 29.23 0.233Issued CP maturing in 41 to 80 days (CP41TO80w) 315 81.44 -112.56 -0.128 32.42 0.419Issued CP maturing in 80 or more days (CP80TOw) 315 118.57 -137.38 -0.075 36.21 0.442Issued ABCP maturing in 1 to 4 days (ABCP1TO4w) 315 79.66 -83.95 -0.116 20.82 0.766Issued ABCP maturing in 21 to 40 days (ABCP21TO40w) 315 90.33 -137.70 -0.267 38.13 0.217Issued ABCP maturing in 41 to 80 days (ABCP41TO80w) 315 101.53 -128.04 -0.100 43.68 0.450Issued ABCP maturing in 80 or more days (ABCP80TOw) 315 182.98 -212.00 -0.002 47.14 0.470Issued FINCP maturing in 10 to 20 days (FINCP10TO20w) 315 172.36 -129.31 -1.291 80.95 0.159Issued FINCP maturing in 21 to 40 days (FINCP21TO40w) 315 161.40 -131.39 -0.336 61.75 0.251Issued FINCP maturing in 41 to 80 days (FINCP41TO80w) 315 221.98 -124.49 -0.266 77.06 0.412Issued FINCP maturing in 80 or more days (FINCP80TOw) 315 218.21 -224.84 0.181 75.04 0.413Issued NONFINCP maturing in 5 to 9 days (NONFINCP5TO9w) 315 204.19 -145.42 -0.015 40.83 0.514Issued NONFINCP maturing in 10 to 20 days (NONFINCP10TO20w) 315 103.09 -113.40 -0.066 37.17 0.703Issued NONFINCP maturing in 41 to 80 days (NONFINCP41TO80w) 315 184.82 -209.95 -0.766 70.11 0.609Issued NONFINCP maturing in 80 or more days (NONFINCP80TOw) 315 300.82 -376.70 -0.514 113.80 0.380

Source: Authors’ calculations based on data from the Weekly Petrolium Status Report, Unemployment Insurance WeeklyClaims Report, and the Board of Governor’s H.15, H.6, and policy statement releases. See Data Appendix for details.

Note: The table reports univariate descriptive statistics of extreme values, central tendency (µ), dispersion (σ), and persistence(ρ) defined as the coefficient of a first-order autoregressive equation (in levels) for each row variable.

119

Figure 2.1: Commercial Paper Outstanding

ALLCPSA ABCPSA FINCPSA

2001 2002 2003 2004 2005 20061200

1300

1400

1500

1600

1700

1800

1900

2000

500

600

700

800

900

1000

1100

1200

FINCPSA NONFINCPSA

2001 2002 2003 2004 2005 2006500

550

600

650

700

750

100

125

150

175

200

225

250

275

300

Source: Authors’ calculations based on data from the Board of Governor’s Commercial Paper release.

Note: The figure plots the weekly level of commercial paper outstanding (in billions of dollars) from January5, 2001 to February 2, 2007. The variables in the top panel are ALLCPSA (left axis), ABCPSA (right axis),and FINCPSA (right axis). The variables in the bottom panel are FINCPSA (left axis) and NONFINCPSA(right axis). Shading indicates NBER business cycle recession dates.

120

Figure 2.2: Money Fund Assets Outstanding

ALLMMMFSA IMMMFSA

YEAR2001 2002 2003 2004 2005 2006

1700

1800

1900

2000

2100

2200

800

900

1000

1100

1200

1300

1400

1500

ALLMMMFSA RMMMFSA

2001 2002 2003 2004 2005 20061700

1800

1900

2000

2100

2200

650

700

750

800

850

900

950

Source: Authors’ calculations based on data from the Board of Governor’s H.6 release.

Note: The figure plots weekly assets at money funds (in billions of dollars) from January 5, 2001 to February 2, 2007. Thevariables in the top panel are: ALLMMFSA (left axis) and IMMFSA (right axis). The variables in the bottom panel are:ALLMMMFSA (left axis) and RMMMFSA (right axis). Shading indicates NBER business cycle recession dates.

121

Figure 2.3: Commercial Paper and the Supply of Credit

3. ABCP

Conduit

4. Broker Dealer

5.MMMF/

MF

6.Monetary Authority

2. Credit

Suppliers

1. Corporations

1. Obligors

Note: The figure maps short term funding flows in the commercial paper market to real economic investment: 1. Corporationswith strong balance sheets issue commercial paper to finance current assets. Obligers and corporations unable to access paperfinancing instead seek credit from loan originators. 2. After granting credit, credit suppliers either finance current assetsdirectly or indirectly by issuing commercial paper. Directly financed paper is issued by the financial institution itself as IOU. 3.Indirect financing requires current assets to be sold to a conduit in a ‘true sale’ with the sponsored conduit then issuing papercollateralized by the recievables held on the balance sheet. 4. Broker dealers place paper by making markets with institutionslooking to invest in low risk short term money market securities. 5. Money funds pool liquidity risk on behalf of investors byinvesting a fraction of their assets in commercial paper while promising to preserve a fixed $1 net asset value. 6. The monetaryauthority conducts daily open market operations in the money market with a subset of broker dealers, the primary governmentsecurities dealers.

122

Figure 2.4: Information Arrival for week = w

Event

AM1

AM2

PM1

PM2

M

FRBNY, 930AM

OIL, 230PM

T

FRBNY

MISS, 7PM

W

FRBNY

TARGET, 215PM

CP, 5PM

Th

CLAIMS, 830AM

FRBNY

MMMF, 430PM

F

FRBNY

FOMC Meeting

Source: Authors’ calculations based on data from the Weekly Petrolium Status Report, Unemployment Insurance WeeklyClaims Report, and the Board of Governor’s H.15, H.6, and policy statement releases. See Data Appendix for details.

Note: The figure is a timeline of information arrival for the days of Monday Jun 28, 2004 to Friday July 2, 2004 based on thevariables included in the benchmark identification.

123

Figure 2.5: FOMC Reaction Function

Shock

toO

IL

0 5 10 15 20 25 30 35 40 45 50-8

-6

-4

-2

0

2

4

6

Shock

toC

LA

IMS

0 5 10 15 20 25 30 35 40 45 50-17.5

-15.0

-12.5

-10.0

-7.5

-5.0

-2.5

0.0

2.5

TARGET

Note: The figure plots impulse responses of the target federal funds rate to OIL (top) and CLAIMS (bottom) from a recursiveVAR model based on the benchmark ordering. The black line is the median of the simulated responses, the blue line represents90% probability bands, and the green line represents 68% probability bands.

124

Figure 2.6: Responses of Jobless Claims, Oil, and the Interest Rate

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

0 5 10 15 20 25 30 35 40 45 50-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

CLAIMS OIL

0 5 10 15 20 25 30 35 40 45 500

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25 30 35 40 45 50-6

-4

-2

0

2

4

6

TARGET MISS

Note: The figure plots clockwise from top left impulse responses of CLAIMS, OIL, MISS, and TARGET after a shock to thetarget federal funds rate based on the benchmark ordering. The black line is the median of the simulated responses, the blueline represents 90% probability bands, and the green line represents 68% probability bands.

125

Figure 2.7: Responses of Money Market Mutual Fund Assets

0 5 10 15 20 25 30 35 40 45 50-0.75

-0.50

-0.25

0.00

0.25

0.50

MMMF

0 5 10 15 20 25 30 35 40 45 50-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

IMMMF

0 5 10 15 20 25 30 35 40 45 50-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

RMMMF

Note: The figure plots from top to bottom impulse responses of MMMF, IMMMF, and RMMMF after a shock to the targetfederal funds rate based on the benchmark ordering. The black line is the median of the simulated responses, the blue linerepresents 90% probability bands, and the green line represents 68% probability bands.

126

Figure 2.8: Responses of Commercial Paper Outstanding

0 5 10 15 20 25 30 35 40 45 50-0.2

0.0

0.2

0.4

0.6

0.8

1.0

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

CP NONFINCP

0 5 10 15 20 25 30 35 40 45 50-0.25

0.00

0.25

0.50

0.75

1.00

0 5 10 15 20 25 30 35 40 45 50-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

FINCP ABCP

Note: The figure plots clockwise from top left impulse responses of CP, NONFINCP, ABCP, and FINCP after a shock to thetarget federal funds rate based on the benchmark ordering. The black line is the median of the simulated responses, the blueline represents 90% probability bands, and the green line represents 68% probability bands.

127

Figure 2.9: Responses by CollateralizationIM

MM

F

0 5 10 15 20 25 30 35 40 45 50-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

0 5 10 15 20 25 30 35 40 45 50-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

RM

MM

F

0 5 10 15 20 25 30 35 40 45 50-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

0 5 10 15 20 25 30 35 40 45 50-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

ABCP UCP

Note: The figure plots impulse responses of ABCP and UCP after a shock to the target federal funds rate based on thebenchmark ordering. The top row includes IMMMF and the bottom RMMMF. The black line is the median of the simulatedresponses, the blue line represents 90% probability bands, and the green line represents 68% probability bands.

128

Figure 2.10: Responses by Credit Rating

0 5 10 15 20 25 30 35 40 45 50-0.25

0.00

0.25

0.50

0.75

1.00

CPT1

0 5 10 15 20 25 30 35 40 45 50-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

CPT2

Note: The figure plots impulse responses of CPT1 (top) and CPT2 (bottom) after a shock to the target federal funds rate basedon the benchmark ordering. The black line is the median of the simulated responses, the blue line represents 90% probabilitybands, and the green line represents 68% probability bands.

129

Figure 2.11: Responses by Intermediation

0 5 10 15 20 25 30 35 40 45 50-3

-2

-1

0

1

2

3

4

DIRPCP

0 5 10 15 20 25 30 35 40 45 50-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

DPCP

Note: The figure plots impulse responses of DIRCP (top) and DPCP (bottom) after a shock to the target federal funds ratebased on the benchmark ordering. The black line is the median of the simulated responses, the blue line represents 90%probability bands, and the green line represents 68% probability bands.

130

Figure 2.12: Responses of Issuance by Maturity - All

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

5

0 5 10 15 20 25 30 35 40 45 50-3

-2

-1

0

1

2

3

4

CP10TO20 CP21TO40

0 5 10 15 20 25 30 35 40 45 50-6

-5

-4

-3

-2

-1

0

1

2

3

0 5 10 15 20 25 30 35 40 45 50-6

-5

-4

-3

-2

-1

0

1

2

CP80TO CP41TO80

Note: The figure plots clockwise from top left impulse responses of commercial paper issued by maturity (short to long) after ashock to the target federal funds rate with paper issuance ordered last. The black line is the median of the simulated responses,the blue line represents 90% probability bands, and the green line represents 68% probability bands.

131

Figure 2.13: Responses of Issuance by Maturity - Asset-backed

0 5 10 15 20 25 30 35 40 45 50-2

-1

0

1

2

3

4

0 5 10 15 20 25 30 35 40 45 50-3

-2

-1

0

1

2

3

4

ABCP1TO4 ABCP21TO40

0 5 10 15 20 25 30 35 40 45 50-5.0

-2.5

0.0

2.5

5.0

0 5 10 15 20 25 30 35 40 45 50-8

-6

-4

-2

0

2

4

ABCP80TO ABCP41TO80


132

Figure 2.14: Responses of Issuance by Maturity - Financial

0 5 10 15 20 25 30 35 40 45 50-4

-2

0

2

4

6

8

0 5 10 15 20 25 30 35 40 45 50-6

-4

-2

0

2

4

6

8

10

FINCP10TO20 FINCP21TO40

0 5 10 15 20 25 30 35 40 45 50-15.0

-12.5

-10.0

-7.5

-5.0

-2.5

0.0

2.5

5.0

7.5

0 5 10 15 20 25 30 35 40 45 50-8

-6

-4

-2

0

2

4

6

FINCP80TO FINCP41TO80


133

Figure 2.15: Responses of Issuance by Maturity - Nonfinancial

0 5 10 15 20 25 30 35 40 45 50-4

-2

0

2

4

6

8

10

0 5 10 15 20 25 30 35 40 45 50-5.0

-2.5

0.0

2.5

5.0

7.5

NONFINCP5TO9 NONFINCP10TO20

0 5 10 15 20 25 30 35 40 45 50-20

-15

-10

-5

0

5

10

15

0 5 10 15 20 25 30 35 40 45 50-10

-8

-6

-4

-2

0

2

4

NONFINCP80TO NONFINCP41TO80


134

Chapter 3

Monetary Policy and the Non-BankFinancial Sector: A Look at Issuers ofAsset-Backed Securities

I Introduction

In the U.S. financial system, term debt markets play a critical role in the supply of credit.

One reason is the regular use of securitization as a means to intermediate savings into

real economic investment.1 Likewise, at the margin, most financial intermediaries are now

market-based. For example, commercial banks are no longer exclusively reliant on insured

deposits and consequently, deposits are no longer the main source of funding for the financial

sector.2

Instead, many financial institutions sponsor pass-through vehicles, or special purpose

vehicles (SPVs), as a way to reduce financing costs by separating business decisions from

funding decisions.3 However, the gains in efficiency by engaging in shadow banking activity

also come with the potential loss of stability due to balance sheet fragility.4 As a result, the

1Securitization is the process of pooling loans and repackaging them into pass-though bonds and com-mercial paper.

2See Woodford (2010) for the description of a ‘market-based’ financial system.3See Gorton and Souleles (2007) for an economic analysis of conduits, also known as special purpose

vehicles.4Bernanke (2010) describes shadow banks as being institutions that are not regulated depository institu-

tions which includes asset-backed conduits, money market mutual funds (MMMFs), and investment banksamong other entities.

135

balance sheet composition of these vehicles, while simple, is a key determinant in the level

of systemic risk in the financial system.5

Although issuers of asset-backed securities (ABS) hold a diverse group of long term assets

(see Table 1), the term of their liabilities fall into two maturities: short and long. In a well

functioning collateralized debt market, intermediaries prefer to finance assets by originating

bonds. With bond financing, the SPV is less susceptible to rollover risk associated with

a constant need to refinance maturing commercial paper. As bonds have similar duration

lengths as the loans backing them, under normal circumstances, the relative cash flows

between the two instruments will offset.

However, if investors were for some reason to become more risk averse about asset-backed

securities, then market conditions may make the cost of financing assets so far out in the

future expensive relative to commercial paper which is traded in the money market. At the

same time, by substituting from bond to paper financing, the SPV increases the duration

gap on its balance sheet.6 As a result, the mix of liabilities can be used as an indicator of

balance sheet risk and by extension, systemic risk in the shadow banking system.

As the initial stages of the monetary transmission mechanism generally take place through

money markets, the actions of the central bank likely impact the markets for asset-backed

instruments in systematically important ways.7 For example, policy changes in the target

interest rate can influence either the perception of risk or the risk-tolerance among investors

in asset-backed credit instruments.8 Furthermore, they can affect the pricing of long term

assets which serve as collateral in asset-backed programs and the bonds issued by these

programs.9

The objective of this paper is to study how issuers of asset-backed securities respond

to changes in monetary policy. We view this to be particularly important in light of the

5Here we define systemic risk as the propagation of default due to fire sales and or rollover risk. For acase study of this mechanisim see Shin (2009).

6Also known as asset liability mismatch or maturity mismatch.7See Friedman and Kuttner (2011).8Also known as the ‘risk-taking channel.’ See Borio and Zhu (2012) for a review.9See Gurkaynak et al. (2005).

136

important role that securitization plays in the extension of credit in the economy. Moreover,

the behavior of issuers of asset-backed securities played a key role in the breakdown of

economic activity during the recent financial crisis.

We find that an anticipated increase in the target for the federal funds rate impacts the

behavior of ABS issuers. In particular, we find commercial paper issuance rises while bond

issuance falls.

However, an increase in paper issuance can also be the result of window dressing in

tough market conditions. In other words, monetary policy might instead be contributing to

SPV activity because it encourages the use of regulatory arbitrage by sponsoring financial

intermediaries. To test this argument, we look at the impact of monetary policy on the

growth in SPV assets. Yet, we do not find a significant response. Therefore, the shift in the

duration gap is likely due to changes in credit market conditions. To that end, monetary

policy does not factor in the sponsoring firm’s decision to use regulatory arbitrage.

This paper contributes to an emerging literature on the transmission of monetary policy

to risk-taking activity in non-bank credit markets. For example, in the market for repurchase

agreements, Bech et al. (2014), Boulware et al. (2014), and D’amico et al. (2013) find sup-

porting evidence of risky behavior by investors in the repo market in response to unexpected

changes in monetary policy. Similarly, in the commercial paper market, Boulware and Reed

(2014) find evidence of duration supply effects from monetary policy. They point out that

the response is a function of the balance sheet strength of issuers of liabilities.10

In terms of theoretical work, Hobjin and Ravenna (2009) attempt to model the impact of

monetary policy on loan securitization. They find a risk channel for conventional monetary

policy operating through the amount of securitized loans held by investors. Nonetheless,

in opposition to our findings, their numerical exercises suggest that monetary policy should

affect the extent of loan securitization.

The remainder of the paper is organized as follows. Section 2 outlines the institutional

10Cahill et al. (2013) find a similar supply effect for the term structure of interest rates following FOMCTreasury purchase announcements.

137

details linking the stance of monetary policy, the market for asset-backed securities, and real

economic investment. Section 3 describes the data and empirical methodology. Section 4

reports the benchmark results. Section 5 concludes.

II Institutional Background

Asset-backed securities represent an important segment of the fixed income market. On the

one hand, ABS and ABCP provide issuers with cheap capital. On the other, they can be used

to deliver income diversification for investors. In this section, we discuss key institutional

details regarding the supply of credit in the U.S. securitized loan market. Figure 1 plots

time series of asset and liability conduit activity. Figure 2 illustrates how financing activity

by issuers of asset-backed securities ultimately extends to the real economy.

II.1 Shadow Banking and the Supply of Credit

When looking at activity among ABS issuers (see Figure 1), growth in conduit assets in-

creased very quickly in the late 1990’s. This behavior coincides with explosive growth in

commercial paper issuance starting in 1996 and peaking in the third quarter of 2001. The

growth in asset-backed paper did not resume until 2005. At the same time asset-backed

bonds steadily increased until 2001, when in a similar fashion as asset-backed paper, bonds

grew at a much faster pace relative to their historical norm.

The stylized flow diagram in Figure 2 links monetary policy, securitization, and the

supply of credit. The diagram starts with obligors, individuals and businesses seeking access

to credit. Examples include consumers applying for a car loan, firms leasing office space,

and entrepreneurs paying startup costs with revolving credit.

Credit suppliers are generally large bank holding companies but also include financial

service subsidiaries of nonfinancial companies such as the General Motors Acceptance Cor-

poration (GMAC) and General Electric Capital. After granting credit to obligors, originators

138

finance a fraction of these assets by sponsoring a conduit. In a “true sale,” loan origina-

tors sell loan obligations to a bankruptcy remote SPV. Using a pass-through vehicle affords

originators lower capital requirements which in turn promotes the further extension of credit.

The balance sheets of ABS issuers are composed of long-term loans repackaged into long

and short-term liabilities of bonds and commercial paper, reflecting their role in maturity

transformation. However, SPVs are passive vehicles. As a result, their liabilities are mainly

intermediated by broker-dealer investment banks. Dealers primarily finance market making

duties by entering into repurchase agreements.

The final agent in Figure 2 is the monetary authority. On a daily basis, trades occur

between the Federal Reserve Bank of New York Open Market Desk and the Primary Gov-

ernment Securities Dealers of the Federal Reserve System (Primary Dealers). These trades

are open market operations.

An open market operation can be either permanent or temporary. For example, in a

permanent liquidity injection, the open market desk purchases bonds from a primary dealer.

That is, the interaction between a Primary Dealer and the open market desk is analogous

to a primary dealer selling asset-backed bonds to a mutual fund.

Furthermore, many of these dealers are subsidiaries of bank holding companies originating

loans and sponsoring asset-backed programs.11 In turn, open market operations can affect

the amount of credit that the dealers and their parent bank holding companies can ultimately

issue. In this manner, monetary policy directly transmits to real activity.12


In the following section we describe the methodology we use to measure monetary policys

impact on the balance sheet of asset-backed issuers. First, we introduce and summarize the

main variables used in our study. Lastly, we present and discuss our empirical model.

11A current list of the primary dealers is maintained by the Federal Reserve Bank of New York at:http://www.newyorkfed.org/markets/pridealers_current.html.

12This transmission channel is explored in more detail by Adrian and Shin (2009).

139

http://www.newyorkfed.org/markets/pridealers_current.html

III.1 Data

Table 1 presents summary statistics for all of the variables used in our analysis and includes

simple measures of dispersion and persistence. Our sample ranges from the second quarter

of 1989 to the third quarter of 2007. Our sample begins at the start of trading in federal

funds futures at the Chicago Board of Trade.

Data on asset-backed issuer balance sheet activity is published in the Federal Reserve

Board of Governors statistical release Z.1 Financial Accounts of the United States. Gross

Domestic Product is published by the Bureau of Economic Analysis. Interest rates are from

the Chicago Board of Trade and are the sum of all within the quarter period changes.

All variables except interest rates have been deflated and transformed into growth rates.13

Detailed descriptions of our data and its sources are listed in the Data Appendix.

III.2 Baseline Model

The following section defines our baseline econometric methodology. In order to measure the

impact of monetary policy on the balance sheet composition of asset-backed security issuers

we estimate three specific to general univariate models:

∆Xq = β0 + γ1∆TARGETq + εq (3.1)

∆Xq = β0 + Σ3j=1βjXq−j + γ1∆TARGETq + εq (3.2)

∆Xq = β0 + Σ3j=1βjXq−j + Σ3

j=1δjYq−j + γ1∆TARGETq + εq (3.3)

where ∆Xq is the quarterly growth rate of the balance sheet variable of interest and Yq is the

quarterly growth rate of real GDP. In each equation, the coefficient of interest is γ1 which

measures the impact effect of a raw target change on the growth of balance sheet activity.

13Each variable was tested for a unit root using the augmented Dickey-Fuller test following the testingprocedure outlined in Enders (2009).

140

III.3 Decomposing Monetary Policy Actions

In our baseline estimation, we do not control for anticipated policy actions. A concern is

that financial markets are forward looking and therefore the results in these regressions may

suffer from endogeneity. One solution is to use a measure of policy expectations in order to

decompose raw target changes into the expected and unexpected moves.

Following Bernanke and Kuttner (2004), we define the unexpected target rate change as

the change in the current month federal funds futures rate the date of the target change

minus the current month’s futures rate the day before:

∆rut = TT−t

(f 0m,t − f 0

m,t−1

)(3.4)

Federal funds futures are based on the average realized effective federal funds rate over

the contract month and therefore must be adjusted by the number of days left in the month.

The fraction in the above equation represents this adjustment. In addition, the superscript

indicates the current month contract for month m.

This measure represents an unbiased measure of monetary policy surprises.14 The ex-

pected target rate change is calculated as the actual target change minus the surprise target

change using the following equation:

∆ret = ∆TARGETt − ∆rut (3.5)

We then add up all of the within period expected and unexpected changes to create a

quarterly expected and unexpected target change. We use these two variables as our isolated

policy instruments in our benchmark regressions. As a result, we estimate augmented base-

line regressions which distinguish between the impact of expected and unexpected changes

in the target funds rate on asset-backed issuer balance sheet growth:

∆Xq = β0 + γe∆req + γu∆ruq + εq (3.6)

14See Piazzesi and Swanson (2008).

141

∆Xq = β0 + Σ3j=1βjXq−j + γe∆req + γu∆ruq + εq (3.7)

∆Xq = β0 + Σ3j=1βjXq−j + Σ3

j=1δjYq−j + γe∆req + γu∆ruq + εq (3.8)

IV Results

In the following section we present our benchmark results. The first section displays results

from our baseline specification. The second section shows results from the augmented base-

line model. Lastly, we conduct a robustness check by looking at the impact of policy on

conduit asset growth.

IV.1 The Reaction of Conduit Liabilities to Changes in the TargetFederal Funds Rate

Tables 3 through 5 present the impact of monetary policy on the growth of asset-backed

bonds, paper, and the mix of liabilities for issuers of asset-backed securities respectively.

We begin with the discussion of asset-backed securities (bonds). The regression results

indicate that an increase in the federal funds target by 100 basis points leads to a two

percentage point drop in the amount of ABS issued. This result is supported in every

specification except for the last two which control for the level of real economic activity.

Table 4 reports the regressions for ABCP. Interestingly, in the specifications which control

for seasonality and real GDP, a monetary contraction is positively correlated with an increase

in commercial paper issuance. In this manner, our results seem to indicate that monetary

contractions induce ABS issuers to switch from issuing long-term liabilities in the form of

ABS to short-term liabilities in the form of ABCP.

142

IV.2 The Reaction of Conduit Liabilities to Federal Funds RateSurprises

Tables 7 through 9 show the results to looking at decompositions between anticipated policy

changes and unanticipated changes in policy. As one might anticipate, expected changes

behave the same way as the target. This could be a consequence of our use of quantities

instead of prices and or studying activity at a lower frequency than one day.

IV.3 Robustness Check: The Reaction of Conduit Assets

As previously mentioned, our findings can also be driven by sponsoring firms engaging in

regulatory arbitrage. We therefore estimate our univariate regressions using conduit asset

growth.

As shown in Tables 9 and 10, the impact of monetary policy on conduit asset growth

insignificant. As a result, we conclude the response of conduit liability growth is the result

of investors responding to the impact of monetary policy on credit market conditions.

V Conclusion

This paper investigates the relationship between monetary policy and risk-taking in the

shadow banking system. Our analysis focuses on issuers of asset-backed securities, firm

sponsored pass-through vehicles used to finance real economic investment via securitization.

We find that an anticipated increase in the target for the federal funds rate impacts the

behavior of ABS issuers. In particular, we find commercial paper issuance rises while bond

issuance falls.

However, this shift in the duration gap is not the result of an increase in the amount of

assets being financed. To that end, monetary policy does not factor in the sponsoring firm’s

decision to engage in regulatory arbitrage.

143

References

Adrian, Tobias, and Hyun Song Shin. 2008. Financial intermediaries, financial stability andmonetary policy. Paper presented at the Federal Reserve Bank of Kansas City Sympo-sium at Jackson Hole, Wyoming, August 21-23.

Bech, Morten, Elizabeth Klee, and Victor Stebunovs. 2012. Arbitrage, liquidity and exit:the repo and federal funds markets before, during, and emerging from the financial crisis.FRB Finance and Economics Discussion Series, 2012-21: 1-54.

Bernanke, Ben S.. 2010. Statement before the financial crisis inquiry commission. 2 Speech.September.

Bernanke, Ben S., and Kenneth Kuttner. 2005. What explains the stock market reaction tofederal reserve policy? Journal of Finance. 60(3), 1221-1257.

Borio, Claudio, and Zhu Haibin. 2012. Capital regulation, risk-taking and monetary policy:a missing link in the transmission mechanism? Journal of Financial Stability, 8(4): 236-251.

Boulware, Karl David, Jun Ma, and Robert Reed. 2014. How do money market conditionsaffect shadow banking activity? evidence from security repurchase agreements. Mimeo.

Boulware, Karl David, and Robert Reed. 2014. Monetary Policy and the Non-bank Finan-cial Sector: A Look at Commercial Paper. Mimeo.

Cahill, Michael E., Stefania D’Amico, Canlin Li, and John S. Sears. 2013. Duration riskversus local supply channel in treasury yields: evidence from the federal reserve’s assetpurchase announcements. FRB Finance and Economics Discussion Series, 2013-35.

D’Amico, Stefania, Roger Fan, and Yuriy Kitsul. 2013. The scarcity value of Treasurycollateral: repo market effects of security-specific supply and demand factors. FederalReserve Bank of Chicago Working Paper, 2013-22.

Enders, Walter. 2009. Applied Econometric Time Series. John Wiley & Sons.

144

Friedman, Benjamin M., and Kenneth N. Kuttner. 2011. Implementation of monetary pol-icy: how do central banks set interest rates? In: Friedman and Woodford (eds.), Handbookof Monetary Economics. Amsterdam, North-Holland.

Gurkaynak, Refet S., Brian Sack, and Eric T. Swanson. 2005. The sensitivity of long-terminterest rates to economic news: evidence and implications for macroeconomic models.American Economic Review. 95(1): 425-436.

Gorton, Gary B., and Nicholas S. Souleles. 2007. Special purpose vehicles and securitization.In The Risks of Financial Institutions, University of Chicago Press, 549-602.

Hobijn, Bart, and Federico Ravenna. 2009. Loan securitization and the monetary trnasmis-sion mechanisim. September, Working Paper.

Piazzesi, Monika, and Eric T. Swanson. 2008. Futures prices and risk adjusted forecasts ofmonetary policy. Journal of Monetary Economics. 55(2): 677-691.

Shin, Hyun Song. 2009. Reflections on northern rock: the bank run that heralded the globalfinancial crisis. Journal of Economic Perspectives. 23(1): 101-19.

Woodford, Michael. 2010. Financial intermediation and macroeconomic analysis. Journalof Economic Perspectives, 24(4): 21-44.

145

Data Appendix

This appendix documents the datasets used in our empirical analysis and is organized by

observation frequency.

Data By Meeting

Federal Funds Futures Prices - The price of the near month federal funds futures contract

is from the Chicago Board of Trade (CBOT) and the monetary policy surprises are published

online by Ken Kuttner at: http://econ.williams.edu/people/knk1/research.

Data By Quarter

Issuers of Asset-Backed Securities - The not seasonally adjusted amounts outstanding

at the end of the period is published online by the Board of Governors in the Z.1 Financial

Accounts of the United States statistical release at: http://www.federalreserve.gov/

releases/z1/current/.

Gross Domestic Product - The seasonally adjusted annual rate of gross domestic prod-

uct is published online by the Bureau of Economic Analysis at: http://www.bea.gov/

national/Index.htm.

146

http://econ.williams.edu/people/knk1/research

http://www.federalreserve.gov/releases/z1/current/

http://www.federalreserve.gov/releases/z1/current/

http://www.bea.gov/national/Index.htm

http://www.bea.gov/national/Index.htm

Table 3.1: The ABS Issuers Balance Sheet - A Snapshot

ASSETS LIABILITIES

Mean Min Max Mean Min Max

Total 1,408 0.22 4,556 Total 1,411 0.22 4,559

1. Single Family Mortgages 584 0 2,315 1. Bonds 1,153 0 3,891

2. Consumer Credit 284 0 679 2. Paper 259 0.22 905

3. Commercial Mortgages 186 0 634

4. Mortgage Backed Securities 151 0 374

5. Securitized (Other) 98 0 380

6. Syndicated (Non-financial) 52 0 302

7. Trade Credit 52 0.22 118

8. Securitized (Non-financial) 46 0 123

9. Multifamily Mortgages 40 0 124

10. Home Equity Loans 20 0.07 88

11. Treasury Securities 13 0 85

Source: Authors’ calculations based on data from the Board of Governors Z.1 Financial Accounts of the United Statesstatistical release. See Data Appendix for details.

Note: The table reports by size (in billions of dollars) descriptive statistics of central tendency and extreme values for thebalance sheet of asset-backed security issuers from Q2 1983 to Q3 2007.

147

Table 3.2: Descriptive Statistics: 1989 Q2 to 2007 Q3

Obs. Max Min µ σ ρ

Panel A: Real Activity

Real GDP (GDPq) 74 1.93 -0.87 0.68 0.64 1.00

Panel B: Cost of Credit

Raw Federal Funds Target Rate Change (∆TARGETq) 74 60.00 -156.00 -6.92 33.44 0.51

Expected Target Change (∆req) 74 61.00 -124.00 -2.51 24.57 0.46

Surprise Target Change (∆ruq ) 74 23.00 -65.00 4.41 13.44 0.37

Panel C: ABS Issuers Balance Sheet

Assets Outstanding (ASSETSq) 74 10.79 -0.47 5.44 11.08 0.998

Bonds Outstanding (ABBq) 74 15.97 -6.34 5.50 11.76 1.007

Commercial Paper Outstanding (ABCPq) 74 24.99 -21.80 4.14 13.11 0.968

Fraction of Bonds to Total Liabilities (MIXq) 74 8.08 -4.58 0.13 1.59 0.97

Source: Authors’ calculations based on data from the Board of Governors Z.1 Financial Accounts of the United Statesstatistical release and the Chicago Board of Trade. See Data Appendix for details.

Note: The table reports univariate descriptive statistics of extreme values, central tendency (µ), dispersion (σ), and persistence(ρ) defined as the coefficient of a first-order autoregressive equation (in levels) for each row variable.

148

Table 3.3: The Impact of Monetary Policy on ABB

1 2 3 4 5 6

Intercept 3.26*** 4.59*** 3.23*** 4.21** 4.47*** 4.59**

(11.45) (2.67) (3.88) (2.20) (4.14) (2.57)

Sum of coefficient on lags 0.01 0.02 -0.01 0.00

(0.04) (0.11) (-0.06) (0.00)

Sum of coefficients on GDP -1.52 -1.47

(-1.55) (-1.49)

∆TARGET -0.02* -0.02** -0.02* -0.02* -0.02 -0.02

(-1.95) (-2.09) (-1.74) (-1.88) (-1.26) (-1.33)

Long-Run Effect -0.02 -0.02 -0.02 -0.02

Quarterly FE N Y N Y N Y

R2 0.11 0.10 0.08 0.06 0.13 0.11

DW 1.23 1.21 1.32 1.30 1.36 1.34

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly growth in outstanding asset-backedbonds for asset-backed issuers on the raw changes in the federal funds target from 1989 Q2 to 2007 Q3. All variables areexpressed in percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-consistent estimates of thestandard errors. The long-run effect is calculated as γ1

1−(Σ3

j=1βj

) . *Statistically significant at the 10 percent level. **Statistically

significant at the 5 percent level. *** Statistically significant at the 1 percent level.

149

Table 3.4: The Impact of Monetary Policy on ABCP

1 2 3 4 5 6

Intercept 5.48*** -10.60** 1.39 -19.17*** -0.39 -18.90***

(6.14) (-2.38) (1.41) (-4.56) (-0.21) (-4.67)

Sum of coefficient on lags 0.70*** 0.83*** 0.71*** 0.83***

(6.67) (9.43) (6.57) (9.18)

Sum of coefficients on GDP 2.27 0.86

(1.00) (0.52)

∆TARGET 0.00 0.02 0.02 0.04* 0.01 0.03*

(0.10) (0.70) (0.60) (1.89) (0.24) (1.65)

Long-Run Effect 0.07 0.24 0.03 0.18


R2 0.01 0.11 0.24 0.49 0.26 0.48

DW 1.03 0.78 1.98 1.80 1.96 1.78

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly growth in outstanding asset-backed commercial paper for asset-backed issuers on the raw changes in the federal funds target from 1989 Q2 to 2007 Q3.All variables are expressed in percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-consistentestimates of the standard errors. The long-run effect is calculated as γ1

1−(Σ3

j=1βj

) . *Statistically significant at the 10 percent

level. **Statistically significant at the 5 percent level. *** Statistically significant at the 1 percent level.

150

Table 3.5: The Impact of Monetary Policy on MIX

1 2 3 4 5 6

Intercept -0.26 2.85*** -0.15 3.18*** 0.07 2.92***

(-1.59) (3.03) (-1.03) (3.78) (0.15) (3.86)

Sum of coefficient on lags 0.45** 0.55*** 0.47*** 0.58***

(2.47) (4.07) (3.83) (4.32)

Sum of coefficients on GDP -0.27 0.01

(-0.46) (0.03)

∆TARGET -0.01 -0.01* -0.01 -0.01 -0.01 -0.01

(-1.13) (-1.70) (-0.85) (-1.40) (-0.67) (1.28)

Long-Run Effect -0.02 -0.02 -0.02 -0.02


R2 0.05 0.18 0.12 0.39 0.16 0.41

DW 1.31 1.10 1.92 1.73 1.86 1.68

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly change in the fraction of outstandingbonds to liabilities for asset-backed issuers on the raw changes in the federal funds target from 1989 Q2 to 2007 Q3. All variablesare expressed in percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-consistent estimatesof the standard errors. The long-run effect is calculated as γ1

1−(Σ3

j=1βj

) . *Statistically significant at the 10 percent level.

**Statistically significant at the 5 percent level. *** Statistically significant at the 1 percent level.

151

Table 3.6: The Impact of Monetary Policy on ABB

1 2 3 4 5 6

Intercept 3.40*** 4.71*** 3.38*** 4.40** 4.55*** 4.70***

(10.28) (2.73) (3.85) (2.30) (4.37) (2.71)

Sum of coefficient on lags 0.00 0.02 -0.01 -0.00

(0.01) (0.08) (-0.95) (-0.01)


(-1.49) (-1.43)

∆req -0.03* -0.04** -0.03* -0.03* -0.03 -0.03

(-1.91) (-2.09) (-1.74) (-1.91) (-1.43) (-1.51)

∆ruq 0.00 0.00 0.00 -0.00 0.01 0.01

(0.29) (0.06) (0.13) (-0.02) (0.56) (0.41)

Long-Run Effect -0.03 -0.03 -0.03 -0.03


R2 0.12 0.11 0.08 0.07 0.14 0.11

DW 1.29 1.27 1.35 1.32 1.40 1.38

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly growth in outstanding asset-backedbonds for asset-backed issuers on the expected and surprise change in the federal funds target from 1989 Q2 to 2007 Q3.All variables are expressed in percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-consistent

estimates of the standard errors. The long-run effect is calculated as γe

1−(Σ3

j=1βj

) . *Statistically significant at the 10 percent

level. **Statistically significant at the 5 percent level. *** Statistically significant at the 1 percent level.

152

Table 3.7: The Impact of Monetary Policy on ABCP

1 2 3 4 5 6

Intercept 4.99 -11.01** 1.29 -19.00*** -0.40 -18.66***

(5.13) (-2.48) (1.27) (-4.53) (-0.21) (-4.53)

Sum of coefficient on lags 0.68 0.81*** 0.69*** 0.81***

(6.86) (8.95) (6.84) (8.96)

Sum of coefficients on GDP 2.14 0.77

(0.91) (0.45)

∆req 0.04 0.06 -0.03 0.05* 0.02 0.05*

(0.98) (1.59) (0.82) (1.90) (0.59) (1.72)

∆ruq -0.09 -0.07 -0.03 0.01 -0.03 0.00

(-1.52) (-1.61) (-0.53) (0.29) (-0.82) (0.12)

Long-Run Effect -0.09 0.26 0.06 0.26


R2 0.00 0.13 0.23 0.48 0.26 0.47

DW 1.08 0.83 1.98 1.81 1.96 1.79

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly growth in outstanding asset-backedcommercial paper for asset-backed issuers on the expected and surprise change in the federal funds target from 1989 Q2 to2007 Q3. All variables are expressed in percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-

consistent estimates of the standard errors. The long-run effect is calculated as γe

1−(Σ3

j=1βj

) . *Statistically significant at the

10 percent level. **Statistically significant at the 5 percent level. *** Statistically significant at the 1 percent level.

153

Table 3.8: The Impact of Monetary Policy on MIX

1 2 3 4 5 6

Intercept -0.15 2.95 -0.06 3.19*** 0.13 2.91***

(-0.76) (3.09) (-0.37) (3.82) (0.33) (3.86)

Sum of coefficient on lags 0.42** 0.52*** 0.44*** 0.55***

(2.47) (3.60) (3.58) (4.26)

Sum of coefficients on GDP -0.24 0.04

(-0.42) (0.08)

∆req -0.02 -0.02* -0.02 -0.02 -0.01 -0.02

(-1.46) (-1.93) (-1.07) (-1.51) (-1.04) (-1.51)

∆ruq 0.01 0.01 0.01 0.00 0.01 0.01

(1.12) (0.85) (1.00) (0.62) (1.29) (0.82)

Long-Run Effect -0.03 -0.04 -0.02 -0.04


R2 0.07 0.20 0.13 0.32 0.27 0.33

DW 1.39 1.18 1.94 1.76 1.90 1.72

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly change in the fraction of outstandingbonds to liabilities for asset-backed issuers on the expected and surprise change in the federal funds target from 1989 Q2 to2007 Q3. All variables are expressed in percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-

consistent estimates of the standard errors. The long-run effect is calculated as γe

1−(Σ3

j=1βj

) . *Statistically significant at the

10 percent level. **Statistically significant at the 5 percent level. *** Statistically significant at the 1 percent level.

154

Table 3.9: The Impact of Monetary Policy on ASSETS

1 2 3 4 5 6

Intercept 3.70*** -0.23 3.71*** -0.59 3.95*** -0.44

(16.17) (-0.18) (4.61) (-0.43) (4.48) (-0.30)

Sum of coefficient on lags -0.00 0.06 0.00 0.08

(-0.01) (0.34) (0.01) (0.43)


(-0.51) (-0.95)

∆TARGET -0.01 -0.00 -0.01 -0.00 -0.01 -0.00

(-1.50) (-0.94) (-1.40) (-0.83) (-1.00) (-0.16)

Long-Run Effect -0.01 0.00 -0.01 0.00


R2 0.01 0.12 0.03 0.09 0.06 0.06

DW 1.24 0.99 1.26 1.05 1.24 1.08

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly growth in asset holdings forasset-backed issuers on the raw changes in the federal funds target from 1989 Q2 to 2007 Q3. All variables are expressedin percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-consistent estimates of the standarderrors. The long-run effect is calculated as γ1

1−(Σ3

j=1βj

) . *Statistically significant at the 10 percent level. **Statistically

significant at the 5 percent level. *** Statistically significant at the 1 percent level.

155

Table 3.10: The Impact of Monetary Policy on ASSETS

1 2 3 4 5 6

Intercept 3.66*** -0.27 3.69*** -0.62 3.94*** -0.45

(14.96) (-0.21) (4.59) (-0.45) (4.48) (-0.30)

Sum of coefficient on lags -0.01 0.06 -0.00 0.07

(-0.03) (0.31) (-0.01) (0.40)

Sum of coefficients on GDP -0/36 -0.55

(-0.55) (-0.97)

∆req -0.00 -0.00 -0.00 -0.00 -0.00 0.00

(-0.61) (-0.16) (-0.58) (-0.02) (-0.33) (0.37)

∆ruq 0.01 -0.01 0.01 -0.01 -0.01 -0.01

(-0.93) (-0.85) (-0.87) (0.85) (-0.76) (-0.61)

Long-Run Effect 0.00 0.00 0.00 0.00


R2 0.01 0.11 0.03 0.14 0.07 0.05

DW 1.23 0.99 1.25 1.40 1.24 1.07

T = 74

Note: The table reports the results from the three baseline OLS regressions of the quarterly growth in asset holdings forasset-backed issuers on the expected and surprise change in the federal funds target from 1989 Q2 to 2007 Q3. All variablesare expressed in percentage terms. Parenthesis contain t-statistics, calculated using heteroskedasticity-consistent estimates

of the standard errors. The long-run effect is calculated as γe

1−(Σ3

j=1βj

) . *Statistically significant at the 10 percent level.

**Statistically significant at the 5 percent level. *** Statistically significant at the 1 percent level.

156

Figure 3.1: Assets and Liabilities of Asset-Backed Issuers Over Time

1989 1991 1993 1995 1997 1999 2001 2003 2005 20070

1000

2000

3000

4000

5000

RASSETS

RABB RABCP

Source: Authors’ calculations based on data from the Board of Governors Z.1 Financial Accounts of the United Statesstatistical release. See Data Appendix for details.

Note: The figure plots quarterly real total financial assets (RASSETS) and liabilities (RABB and RABCP) outstanding forissuers of asset-backed securities (in billions of dollars) from Q2 1989 to Q3 2007. Shading indicates NBER business cyclerecession dates.

157

Figure 3.2: Securitization and the Supply of Credit

3. SPV

4. Broker Dealer

5.MMMF/

MF

6.Monetary Authority

2. Credit

Suppliers

1. Obligors

Note: The figure maps short term funding flows in the market for asset-backed securities to real economic investment: 1.Obligers and corporations seek credit from loan originators. 2. After granting credit, credit suppliers finance a percentage ofassets indirectly by issuing asset-backed securities. 3. Indirect financing requires assets to be sold to a conduit in a ‘true sale’with the sponsored conduit then issuing paper and or bonds collateralized by the recievables held on the conduits balance sheet.4. Broker dealers place securitized bonds and paper by making markets with institutions looking to diversify their fixed incomesecurity holdings. 5. Money funds and mutual funds pool liquidity risk on behalf of investors by investing a fraction of theirassets in asset-backed securities. 6. The monetary authority conducts daily open market operations in the bond and moneymarket with a subset of broker dealers, the primary government securities dealers.

158

Conclusion

The conclusions reached in this dissertation have important implications for monetary pol-

icy. The first essay shows that monetary policy contributes to rollover risk for the primary

government dealers. In particular, there is a shift from long to short repurchase agreement

activity. As a result, we argue macro prudential policy should be an important concern for

monetary authorities.

The second and third essays present similar findings while using stronger identification

methods than in the first essay. Specifically, in the second essay we find evidence of a liquidity

risk channel of monetary transmission operating through the total supply of commercial

paper. In other words a monetary tightening is associated with a reduction in liquidity in

the money market and a temporary reduction in paper issuance for firms with weak balance

sheets. Furthermore, we find evidence of a duration effect where investors ration credit to

the weakest borrowers by maturity. In the third essay we find systematic policy impacts

off-balance sheet non-bank credit activity significantly. Simillarly, there is a duration shift

in issuance by asset-backed security issuers from bonds to commercial paper. As a result,

the gap on their balance sheet between long term assets and short term liabilities increases

thereby making them more susceptible to rollover risk.

In summary, we find supporting evidence for a liquidity risk channel of monetary trans-

mission and an increase in rollover risk for non-bank credit issuers. As a result we conclude,

monetary policy contributes to systemic risk in the financial system and wholesale finding

markets are an important link between monetary policy and financial stability.

159

THREE ESSAYS ON THE TRANSMISSION...

Documents

Transcript of THREE ESSAYS ON THE TRANSMISSION...